Blog

From National Champion Archer to Celebrity Fitness Trainer & Entrepreneur, Laura Wilson Expands NATURAL PILATES on the West Coast and Introduces a New Studio to the East Coast – Yahoo Finance

Expansion Includes New Studios in Sherman Oaks, CA and New York City – Soho
LOS ANGELES, Aug. 1, 2022 /PRNewswire/ — Laura Wilson is one of the most sought-after fitness experts in the country who specializes in Pilates. In 2005 she launched Natural Pilates with a studio in Beverly Hills and for 10 years grew her clientele, eventually expanding to additional locations in Brentwood (2017) and West Hollywood (2019). Today, Wilson announces the opening of two more studios, one in Sherman Oaks, a Los Angeles neighborhood in the San Fernando Valley and the other in SoHo, a New York City neighborhood in Lower Manhattan. The East coast location will be the first for the brand outside of the Southern California area.
The Natural Pilates training method is a form-focused athletic approach to classical Pilates helping clients to increase strength, improve flexibility, and gain mindfulness. Reformer classes and private sessions are available at all locations, with Brentwood and West Hollywood offering outdoor options. Through Natural Pilates TV, access to Pilates can be anytime, anywhere with over 250+ mat/reformer workouts for every fitness level. Fitness props can be purchased from the online store and for those interested in taking their practice further, Natural Pilates offers a Teacher Training Program and educational workshops.
Wilson’s success has been one of discipline and focus that began in her native country of Romania where she grew up a gymnast, dancer and member of Romania’s National Archery team winning two national titles. As a young, outstanding athlete, Wilson always displayed an affinity for discipline and setting high goals for herself.  While attending college in Romania, Wilson worked in cosmetology and applied for an inter-exchange student program where she would travel to the U.S and work for three months. Wilson was accepted and arrived in America in 2002 at the age of 21 with $100 in her pocket. Taking notice of the lifestyle and opportunities presented to her, she decided to stay and apply for college classes in Los Angeles. Wilson took a job as a massage therapist and began learning about healing and the physical body. Soon after, she discovered the movement of Pilates and viewed it as the perfect fit for strengthening, stretching, and healing the body, and decided this was the career she wanted.
It is through humble beginnings that Wilson established a Pilates empire in both Los Angeles and online. She was able to fully self-finance the first studio and then eventually the others followed. In the pandemic, Wilson safely maneuvered her business and clientele by moving two studios outdoors, revising programming schedules, setting up Zoom classes & sending goodie bags of props (bands, weights, fitness circles, etc.) to attendees while not having to lay off any staff.  At the same time the shutdown was happening, Wilson immediately launched Natural Pilates TV, an online subscription that provides workout routines even when livestream class times may not fit a person’s schedule.  She even decided to sell all the reformers from the studio to clients for their workouts.  Like so many business owners, Wilson was doing it all…. filming, editing, creating class content, and constantly reinventing her spaces to keep everyone safe.
It is through Wilson’s numerous certifications (Stott Pilates-2005, Gyrotonic-2008, Vinyasa Yoga-2011, Resistance Training Specialist and Muscle Activation Technique Specialist-2012), training and extensive knowledge of the body that enables her to provide the most comprehensive workouts tailored to each person’s individual needs. Together with her team, Wilson trains teens, adults, and even active seniors as well as athletes and anyone recovering from injuries.
Both Natural Pilates-Sherman Oaks (15222 Ventura Boulevard) and Natural Pilates-SoHo (206 Spring Street) are scheduled to open September 2022. For more information on studio openings classes, the brand, etc. please visit www.NaturalPilates.com.
About Natural Pilates
Natural Pilates was founded in 2005 by Laura Wilson, a national champion archer from Romania who moved to Los Angeles and discovered Pilates while working as a massage therapist. She found it to be the perfect combination of mental focus and strength training while helping people to feel better.  The brand features Pilates on Demand, teacher training programs and studio classes and private sessions in the Beverly Hills, Brentwood, West Hollywood, and Sherman Oaks areas of Southern California in addition to New York City’s SoHo neighborhood.   Natural Pilates’ training method is a form-focused athletic approach to classical Pilates helping clients to increase strength, improve flexibility and gain mindfulness. www.naturalpilates.com
Press Contact:
Jennifer Lopez
Clover PR
310-864-8633
[email protected]
View original content to download multimedia:https://www.prnewswire.com/news-releases/from-national-champion-archer-to-celebrity-fitness-trainer–entrepreneur-laura-wilson-expands-natural-pilates-on-the-west-coast-and-introduces-a-new-studio-to-the-east-coast-301596966.html
SOURCE Natural Pilates
Gabrielle Union is strong and sculpted in a nude photo she posted to Instagram. The actress, 49, has been embracing her strength and mental health goals lately.
Kelsea Ballerini flaunted her epic abs in a crop top while dancing on TikTok. She enjoys ballet and walking to stay active while conserving her energy each day.
Want to get in a full workout without leaving the house? Take a look at these workout tools that are a perfect fit for any home gym enthusiast.
"The U.S. housing market continues to move toward more evenly balanced supply and demand compared to the 2021 frenzy," Danielle Hale, Realtor.com's chief economist, said in a statement. "Our July data shows elevated mortgage rates left many buyers tightening their budgets and sellers responding with price reductions, while home shoppers who kept searching saw more available options." Meanwhile, "new listings declined in July, suggesting that some prospective sellers are wondering what recent market shifts mean for their plans to list," Hale said.
Helios Technologies (HLIO) delivered earnings and revenue surprises of 0% and 0.44%, respectively, for the quarter ended June 2022. Do the numbers hold clues to what lies ahead for the stock?
MicroStrategy's new CEO said Tuesday that owning bitcoin was the best treasury strategy for the company.
Startups have to brace for a serious market downturn–but it's a great time to bootstrap a new business.
Canada's main stock index rose on Wednesday, with technology stocks leading gains, after data showed a slower-than-expected rise in U.S. inflation last month. At 10:06 a.m. ET (1406 GMT), the Toronto Stock Exchange's S&P/TSX composite index was up 194.76 points, or 0.99%, at 19,773.06. Technology stocks jumped 2.7% as Converge Technology Solutions surged over 12% to lead gains.
"In the (hopefully unlikely) event that Twitter forces this deal to close *and* some equity partners don't come through, it is important to avoid an emergency sale of Tesla stock," Musk said.
A Coinbase shareholder is seeking damages from nine company executives and board members on behalf of the exchange for its allegedly misleading public listing. This comes as Coinbase was hit with another class action lawsuit last week on suspicion it knew it listed digital assets which should have been registered with the SEC. CoinDesk Global Policy and Regulation Managing Editor Nikhilesh De discusses the details.
Trimble's (TRMB) second-quarter results are driven by solid momentum across buildings & infrastructure, and resources & utilities segments.
by Mary-Jane McQuillen, Head of ESG, ClearBridge Investments
Novavax widely missed second-quarter forecasts Tuesday, leading the Covid vaccine maker to slash its guidance and NVAX stock to crash.
NASDAQ's Market Services unit generates the most revenue and profits, but Corporate Platforms and Investment Intelligence have fastest profit growth.
The broadcasting company benefited from higher ratings, political advertising revenue and growth at its ad-supported streaming service Tubi.
(Bloomberg) — Chinese e-commerce giant Alibaba Group Holding Ltd. let go of 9,241 employees in the three months to June, according to the company’s latest filing.Most Read from BloombergMusk Sells Another $6.9 Billion of Tesla Ahead of Twitter TrialUK Plans for Blackouts in January in Emergency Energy PlanUS Inflation Runs Cooler Than Forecast, Easing Pressure on FedRussia Is Scouring the Globe for Weapons to Use Against UkraineDozens in China Infected With New ‘Langya’ Virus Carried by ShrewsT
The Dow Jones Industrial Average surged 460 points on cool CPI inflation data. Elon Musk sold nearly $7 billion worth of Tesla stock.
IBD Stock Of The Day: Range Resources flashed an early entry, rebounding from its 50-day line and breaking a downtrend as natural gas prices rise.
With the Federal Reserve boosting its benchmark federal funds rate by 0.75% in June and July, and with no sign of letting up on rate hikes, prices have gone up for a wide variety of goods and services. BNPL options are especially popular with digital shoppers, who only make a modest initial payment up front and pay the remaining balance on scheduled due dates spelled out during the buying process. A separate study by AC Cutts & Associates notes that 36.9% of BNPL consumers use the payment option strictly to avoid paying credit card interest rates.
Groupon's (GRPN) second-quarter 2022 results suffered from a declining active customer base.

source

Live election results: Omar, McCollum fend off challengers – MPR News

Reps. Ilhan Omar and Betty McCollum have fended off primary challenges. The matchup between DFL Gov. Tim Walz and Republican Scott Jensen is set. Both won their party nominations for governor in today's primary election.
Results for the Hennepin County sheriff and attorney races are available from the Minnesota Secretary of State.
MPR News is made by Members. Gifts from individuals fuel the programs that you and your neighbors rely on. Donate today to power news, analysis, and community conversations for all.

source

Sweetwater opens Gear Exchange marketplace for used instruments – Digital Commerce 360

Get access to essential strategic content, in-depth reports, industry intelligence, and exclusive data.
Get access to essential strategic content, in-depth reports, industry intelligence, and exclusive data.
Get access to essential strategic content, in-depth reports, industry intelligence, and exclusive data.
Get access to essential strategic content, in-depth reports, industry intelligence, and exclusive data.



Forgot your password?
Sweetwater has opened a standalone marketplace, Gear Exchange, for used musical instruments and equipment. But one of the main goals of the marketplace is to drive sales of new instruments on the company’s main website.
“It’s a growth initiative for sure,” said Andy Rossi, vice president of Gear Exchange. “And it’s sort of parallel world to what sweetwater.com is known for, which is new products.”
The key to Gear Exchange is that sellers can avoid any fees associated with selling their equipment if they take their payment in the form of a gift card for Sweetwater.
“You can hold on to 100% of what you sell that product for, and you put it into a Sweetwater gift card and then use it on SweetWater.com for any of the millions of products that are on that site,” Rossi said. “That is boosting your buying power. That is the most cost-effective way to sell and buy. Sweetwater is an active new product seller connected to an active used seller marketplace. That connection is unique.”
Rossi said the idea for the marketplace came from feedback among Sweetwater customers. Those customers often asked about trading their older instruments for better prices. Sweetwater also operated an online bulletin board that connected buyers and sellers of used instruments and equipment. Gear Exchange takes things a step further, facilitating the sale of used equipment and offering the no-fee/gift card payout to ensure the money from used-equipment sales is used for new equipment purchases.
 
Rossi said musicians are often interested in trading up for higher-quality gear.
“Every guitar player has a mirror that that he or she looks in, because that’s part of it. Seeing yourself in that way. It’s part of that dream,” he said. “And that dream leads to you wanting to buy or sell or both to get the dream instrument. It’s always tough to just spend money on the things you want, the things you dream about, as opposed to the things you need, like milk and bread. So the majority of the sellers that you see on a place like Gear Exchange are active music makers who are interested in moving upward in quality or specifications.”
Sweetwater, based in Fort Wayne, Indiana, is ranked No. 71 in the Digital Commerce 360 Top 100 database of online retailers.
Besides the gift-card option, Gear Exchange is much like any other niche marketplace.
Sellers can create their own customized storefront, outline policies for returns and add a biography page. Sellers’ listings can include photos of the instruments and gear they no longer need or want. 
There are no listing fees on Gear Exchange. If a seller doesn’t want a gift card, Gear Exchange collects 5% of the transaction and a 2.5% bank-processing fee. There are no other fees.
Gear Exchange, like Sweetwater, runs on a home-built platform.
“We have our own engineers, our own web developers,” Rossi said. “We have our own fraud detection, our own tech support, our own customer support.”
Customer support and in-house expertise is arguably Sweetwater’s greatest competitive strength. Many of its customer service and sales staff have studied audio engineering or have other music-related degrees. Additionally, nearly every new Sweetwater employee must complete an intense 13-week training program called Sweetwater University. More than 80 industry experts run 300 classes on subjects including digital audio and modern production software to analog circuitry and acoustics.
“There’s a real culture here of helping dreamers fulfill their dream,” Rossi said. “That sounds pretty salesy, I know, but it actually is real. And that’s really what we do.”
With Gear Exchange, the company now has another vehicle to help musicians chase their dreams and help Sweetwater drive revenue.
Stay on top of the latest developments in the ecommerce industry. Sign up for a complimentary subscription to Digital Commerce 360 Retail News.
Follow us on LinkedIn, Twitter and Facebook. Be the first to know when Digital Commerce 360 publishes news content.
Copyright © 2022 Digital Commerce 360 | Vertical Web Media LLC
Copyright © 2022 Digital Commerce 360 | Vertical Web Media LLC

source

Martha King earns Master Pilot Award — General Aviation News – General Aviation News

General Aviation News
Because flying is cool
By · · Leave a Comment
During the 2022 EAA AirVenture Oshkosh, Martha King, one of the co-founders of King Schools, was presented with the FAA’s Wright Brothers Master Pilot Award.
Acting FAA Administrator Billy Nolen presented Martha with her commemorative plaque.
“This isn’t in my prepared remarks, but I just have to say it,” Nolen said during the ceremony. “After thanking Martha, I will say that her tapes are the reason I am standing here. When I became an airline pilot it was the King tapes that I used on my flight engineer written that I had to take and be able to pass. And I used her in my ATP training as well. You are a great contributor to my successful journey and the arc of my career. Thank you again, Martha, for everything you are doing.”
In his remarks during a presentation the day before, Aviation Safety Airworthiness FAASTeam Program Manager Ken Kelley noted that “Martha and her fellow Master Pilot John used technology and clear, simple, and fun teaching to make aviation knowledge more accessible to pilots. They have sold more than 700,000 video-based knowledge courses and have had millions of short-subject video views of their ‘Take-off’ courses.”
John received his FAA Master Pilot Award in 2014.
A pilot since 1969, Martha has all seven flight instructor ratings — Airplane Single and Multi-engine, Rotorcraft Helicopter and Gyroplane, Instrument Airplane and Helicopter, Glider — as well as her Advanced and Instrument Ground Instructor ratings. She holds type ratings in the Falcon 10, LearJet, Citation 500, Eclipse 500S, and Citation Mustang 510S.
“As a pilot and a teacher of learning pilots, I have had a meaningful and rewarding life,” Martha said. “When you help people fly, you help them towards success not only as a pilot, but in life. You help them develop aptitudes that allow them to more fully enjoy life and be contributors to society.”
The Wright Brothers Master Pilot Award is the most prestigious award the FAA issues to pilots certificated under Title 14 of the Code of Federal Regulations (14 CFR) part 61. It recognizes pilots who have exhibited professionalism, skill, and aviation expertise for at least 50 years. The Master Pilots “Roll of Honor” is at FAASafety.gov.
Join 110,000 readers each month and get the latest news and entertainment from the world of general aviation direct to your inbox, daily. Sign up here.
Your email address will not be published.








This site uses Akismet to reduce spam. Learn how your comment data is processed.
© 2022 Flyer Media, Inc. All rights reserved. Privacy Policy

source

37th Annual Stellar Awards: Legendary Producer Aaron Lindsey Receives the 'Aretha Franklin Icon' Award – BET

August 7, 2022
8:25 PM
The 37th annual Stellar Awards honored legendary music producer Aaron Lindsey with the “Aretha Franklin Icon” award Presented by Aflac. The prestigious accolade is given to individuals in the music industry for their outstanding contributions to Gospel music.
Grammy Award-winning singer-songwriter Jonathan McReynolds introduced Lindsey as this year’s award recipient before going on to share kind words about the musician who he described as “a key contributor to the greatest tracks of the last twenty years.”
“This versatile writing and production wiz is known as Gospel’s secret weapon. In fact, you can say that if were not for Aaron some of gospel’s best and brightest, never would’ve made it, they never would’ve made,” McReynold shared, putting his own spin to Marvin Sapp’s 2007 smash single “Never Would’ve Made It” produced by the honoree.
RELATED:  37th Stellar Gospel Music Awards Nominations Announced
Later, singers Brian Courteny Wilson, Koryn Hawthrone, and Sapp took the stage with an unforgettable performance of their popular Lindsey-produced tracks, including “Worth Fighting For,” “Speak The Name,” and “Never Would Have Made It,” respectively.
Following the musical tribute, Lindsey humbly accepted his award. In his speech, the 50-year-old shared, “I could probably find a way to creative a narrative that suggests that in some way I had the ability create this and do this by myself, but I’m far too smart for that cause I know had it not been for the lord that was on my side.” He added, “And I also realize that it’s not by might not by power but is by God’s spirit that we do this music that we do.”
Lindsey  has been in the industry for over two decades but he’s been surrounded by music all his life as the son of a Minister of Music and choir director. He’s worked with many of the greats in gospel and mainstream music including Smokie Norful, Tasha Cobbs, Donnie McClurkin, Fred Hammond, Mary Mary and so many more. He also has produced and written for renowned R&B artists such as India Arie, Kenny Lattimore, Chrisette Michelle, and many others.
The Stellar Gospel Music Awards celebrated Gospel music with an uplifting weekend of events, concluding with the live-taped ceremony on Saturday, July 16, at Cobb Energy Performing Arts Centre in Atlanta, Georgia. The show was hosted by award winning singers, Jekalyn Carr and Kierra Sheard, and premiered on Sunday, August 7 at 8 PM ET/PT on BET, BET Her, and BET International.
Trending Now
By clicking subscribe, I agree to receive newsletters, marketing communications, updates, special offers (including partner offers), and other information from BET and the Paramount family of companies. For more information about our data practices, consult our Privacy Policy.
© 2022 BET Interactive, LLC, a wholly owned subsidiary of Black Entertainment Television LLC. All rights reserved.

source

New Songs: Pixies, Dungen, Lambchop, Mykki Blanco & more – Brooklyn Vegan

So many artists, so many songs, so little time. Each week we review a handful of new albums (of all genres), round up even more new music that we'd call "indie," and talk about what metal is coming out. We post music news, track premieres, and more all day. We update a playlist weekly of some of our current favorite tracks. Here's a daily roundup with a bunch of interesting, newly released songs in one place.
PIXIES – "VAULT OF HEAVEN"
Pixies take things into Spaghetti Western territory on this new single from their forthcoming album Doggerel (out September 30) and that goes double for the video.
—
LAMBCHOP – "SO THERE"
“‘So There,’ in my head, addresses the idea of showing up, being there for your friends and for things that we believe to be right and true," says Kurt Wagner of this new song from Lambchop's upcoming The Bible. "But also wondering if that alone will ever be enough. I was reflecting on this during the time I was immersed in the scorching Minneapolis summer of 2021. Stuck in traffic staring at a Grateful Dead bumper sticker, parked daily under a highway billboard sign with ‘Cloudy Forever’ sprayed on its pole. Flies in the vehicle, an underlying sense of dread on the streets, waiting for the sun to burn away the morning’s haze.” The song comes with a somewhat unsettling video by Isaac Gale who says it was an experiment in learing new VFX software. "Sometimes the best way to learn a new software is to first learn how to break it. So, we decided to see what would happen if we simply flew a camera through Kurt’s head to see what goes on in there.”
—
DUNGEN – “OM DET FINNS NÅGOT SOM DU VILL FRÅGA MIG"
Gustav Estjes says this new Dungen track is "about holding secrets and truth on a personal basis." It's in Swedish so we'll have to take is word on that, but musically this is another soaring psych gem. New album En Är För Mycket och Tusen Aldrig Nog is out October 7 via Mexican Summer.
—
THE BLACK ANGELS – "WITHOUT A TRACE"
"We have always said that if you can rob a bank to our music then we are in the right ballpark," say The Black Angels of the latest song to be released from their upcoming Wilderness of Mirrors. "And while we don't condone robbing a bank – the idea alone creates an anticipatable, adrenaline inducing soundtrack for your mind." Plan your entirely theoretical 4 minute, 8 second heist around this one now.
—
ISAAC BROCK – "I HEARD IT THROUGH THE GRAPEVINE" (NORMAN WHITFIELD & BARRETT STRONG COVER)
The Modest Mouse frontman's menacing cover of the classic track made famous by Marvin Gaye is from the soundtrack to Claydream.
—
DOKI DOKI – "LIAR LIAR"
Doki Doki, the new Asian Man Records-signed indie-punk supergroup with members of Grumpster, Dog Party, Small Crush, and The Moore Family Band/Get Married, have just followed their debut single "I Was Killing It Man" with second single "Liar Liar." It takes a childhood playground chant and turns it into an indie-punk ripper.
—
SPIRITWORLD – "MOONLIT TORTURE" (ft. INTEGRITY'S DWID HELLION)
Stu Folsom's metallic hardcore band SpiritWorld have followed their debut LP Pagan Rhythms with a new single, "Moonlit Torture," and it's truly pulverizing stuff that sounds like Slayer at their punkest, and it features Integrity vocalist Dwid Hellion.
—
HIGH VIS – "TRAUMA BONDS"
Britpop/baggy-infused punks High Vis have shared the latest taste of their upcoming album Blending, and this one was inspired by the tragic suicide of one of the band's friends during lockdown. It's more somber than the previous singles, but as vocalist Graham Sayle wails "Are we still lucky to be here?", it sounds hopeful too.
—
SECRET MACHINES – "THE FIRE IS WAITING" FT TONY VISCONTI
Secret Machines have announced a digital reissue of their self-titled 2008 record, which has been remastered by Simon Scott (Slowdive). With the news they've shared a revamped version of "The Fire Is Waiting," featuring Tony Visconti on recorder.
—
HEAVEN FOR REAL – "ENERGY BAR"
Halifax band Heaven for Real have shared the title track to their new album Energy Bar. The song rides between janglepop and a krautrock groove. "'Can You Believe It?' is one of the great intangible/rhetorical questions," says the band's Mark Grundy. "The asker could be talking about a world-ending climate disaster or a BOGO sale – it is the title track as we felt like it encapsulated a lot of the asking and the feelings associated with this record."
—
HAGOP TCHAPARIAN – "ROUND"
British-Armenian producer Hagop Tchaparian has announced his debut album, Bolts, which will be out October 21 via Kieran Hebden’s (Four Tet) Text Records. Mixing techno with field recordings, he makes a unique brand of blissed out dance music as you can hear on the album's new single, "Round."
—
MARLON WILLIAMS – "EASY DOES IT"
Marlon Williams' new album, My Boy, is out September 9 via Dead Oceans and he's just released this breezy track from it. “It happened naturally,” Williams says of the song. “I was listening to more steely, New Romantic stuff, like Duran Duran, John Grant, Perfume Genius, the Bee Gees. All those things fed into the machine.”
—
ZANNIE – "GET THAT STAR"
Zannie's new album, How Do I Get That Star, is out next week via Kill Rock Stars and they've just shared the quasi title track, a pretty, skyrocketing torch song thick with harmonies and slide guitar.
—
MYKKI BLANCO – "STEPS" FT. MNEK & SAUL WILLIAMS
"I think a song like 'Steps' is a testament to what can happen when three very unique artists are given time to create a meditation together – to respond to one another when the moment is right – almost like the lines of a haiku – maybe 'Steps' is a prayer unique in its ability to transcend and transform and provide peace without necessarily providing an answer," Mykki Blanco says, continuing, "There are certain people I have waited to collaborate with my entire career and Saul Williams is one of those people." It's from their new album Stay Close To Music.
—
SIGH – "SATSUI"
Japanese avant-metal vets Sigh have shared a new track off their upcoming album Shiki (due 8/26 via Peaceville), and it's a totally nuts sounding sound that finds Sigh's Mirai Kawashima offering up views on the death penalty.
—
EZRA FURMAN – "POOR GIRL A LONG WAY FROM HEAVEN"
The latest single from Ezra Furman's new album All of Us Flames is "Poor Girl A Long Way From Heaven." "The spiritual life ain’t all pious platitudes," Ezra says. “This song is about how weird it gets, when you’re in love with the Source of Being and She’s not texting you back. Ever since it hit me that I was never going to be loved and accepted on the scale of my pop star heroes, me and my bandmates have started to work on a different vision of pop, one more our own, one that gestures at the stranger truths of the human mind. Here we are in thrall to verbally adventurous nineties music like Bjork and Beck and the Silver Jews and them kinda non-linear geniuses.”
—
ORBITS (THE SIDEKICKS, SIGNALS MIDWEST, etc) – FEVER MUSEUM
Orbits is the project of Toby Reif (the Sidekicks) and Maxwell Stern (Signals Midwest, Timeshares, Meridian), and their debut record Fever Museum was made with drummers Adam Beck (Into It. Over It., Sincere Engineer) and Eric Slick (Dr. Dog, Waxahatchee). It treks through punk, indie rock, emo, power pop, and more, making use of crunchy electric guitars, soft acoustic ones, buzzing synths, and more, as the band deliver their catchy, quirky melodies.
—
RAZOR – "FLAMES OF HATRED"
Canadian thrash/speed metal legends Razor have announced their first new album in 25 years, and you can read more about lead single "Flames of Hatred" here.
—
OPEN MIKE EAGLE – "I'LL FIGHT YOU" (prod. DIAMOND D)
Open Mike Eagle has announced a new project, Component System With The Auto Reverse, and new single "I'll Fight You" was produced by Diamond D of DITC. Read more about it here.
—
DANGER MOUSE AND BLACK THOUGHT – "STRANGERS" (FEATURING A$AP ROCKY AND RUN THE JEWELS)
"Strangers" sees Black Thought, A$AP Rocky, El-P, and Killer Mike spit tight verses over Danger Mouse's busy, garage-y production–a departure from the smoother R&B sound we've heard in earlier singles. Read more about it here.
—
BLONDIE – "MR. SIGHTSEER"
"Mr. Sightseer" is a previously-unreleased rarity from Blondie's upcoming box set Blondie: Against The Odds 1974-1982. The song is purely fun, a fairly simple bedroom demo recorded by Debbie Harry and Chris Stein in their house in 1978. Read more about it here.
—
JID – "DANCE NOW"
Today, JID unveiled the album art for his upcoming third record The Forever Story, along with the release of third single "Dance Now." The song features contributions by Kenny Mason and Foushee, and was produced by Christo. "Dance Now" is a sendup to Atlanta, with hard hitting verses, super smooth beats, and a harmony-fueled chorus. Read more about it here.
—
JAKE TITTLE – "NOSE BLEEDS" FT. ILLUMINATI HOTTIES
Los Angeles-based singer and songwriter Jake Tittle announced a new album, Blood Offering, and shared a new single featuring Sarah Tudzin aka Illuminati Hotties, which you can read more about here.
—
Looking for even more new songs? Browse the New Songs archive.

source

Auxetic structures used in kinesiology tapes can improve form-fitting and personalization | Scientific Reports – Nature.com

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.
Advertisement
Scientific Reports volume 12, Article number: 13509 (2022)
219 Accesses
Metrics details
Each year 65% of young athletes and 25% of physically active adults suffer from at least one musculoskeletal injury that prevents them from continuing with physical activity, negatively influencing their physical and mental well-being. The treatment of musculoskeletal injuries with the adhesive elastic kinesiology tape (KT) decreases the recovery time. Patients can thus recommence physical exercise earlier. Here, a novel KT based on auxetic structures is proposed to simplify the application procedure and allow personalization. This novel KT exploits the form-fitting property of auxetics as well as their ability to simultaneously expand in two perpendicular directions when stretched. The auxetic contribution is tuned by optimizing the structure design using analytical equations and experimental measurements. A reentrant honeycomb topology is selected to demonstrate the validity of the proposed approach. Prototypes of auxetic KT to treat general elbow pains and muscle tenseness in the forearm are developed.
An injury prevents professional athletes and amateurs alike from their regular sports activities, thus negatively altering their quality of life and habits. Studies show that 65% of young athletes suffer from sports-related injuries each year1, and that among physically active adults, 25% of them suffered from at least one musculoskeletal injury over a one-year period2. It is also reported that 75% of the men and 68% of the women were unable to continue exercising due to the injury2. Such interruptions can negatively affect both the physical and mental status of an athlete1, involving fitness decrease3 and lowered self-esteem4. The latter is especially the case for athletes with a strong athletic identity5, for which isolation from their team members in combination with reduced performance can worsen their mental well-being1 and consequently even lead to depression6.
To speed up the recovery process, a combination of therapies and exercises is applied7,8. Nowadays, the recovery most often involves an adjunctive treatment in which a flexible and adhesive tape, called kinesiology tape (KT), is used. Invented in the 1970s by Kenzo Kase and internationalized by Japanese volleyball players during the Seoul 1988 Summer Olympics9, in its current design, the KT is realized as a cotton fabric wrapped around elastic polymer fibers10. It allows patients to recommence their training at an earlier stage compared, for example, to the standard inelastic tape, as it does not restrict movements7,8. The tape is applied to the athlete/patient’s skin at the location of the injured muscles, ligaments, or joints, which causes the compression of the surrounding blood and lymphatic vessels, as well as the pain receptors (Fig. 1a), to ultimately increase pain and produce swelling11. However, the application of the KT lifts the skin around the inflammation, separating it from the tissue below and thus reducing the compression on blood and lymphatic vessels as well as the pain receptors (Fig. 1b)12,13. This effect is propagated to the injured muscle, joint or ligament, and surrounding tissue to facilitate the natural healing process9 and post-injury treatment14,15, enhance muscle activity16,17 and pain modulation7,8,14, provide neuro-sensory feedback to improve proprioception7,18 and assist in injury prevention18,19. The KT adhesion at the injury location occurs due to an acrylic glue7,10,20, which is applied underneath the tape in a wave-like pattern (Fig. 1c) for strong attachment21. The beneficial effect of the KT arises because the tape is first pre-stretched on average by a 25–75% extension and then applied to the skin10,20, which, as a result, folds up in a wave-like pattern (Fig. 1d).
Existing KT (a) The compression on the skin, fascia, and muscle due to an injury. Image inspired by22,23,24. (b) The positive decompressive effect generated by the KT application. A one-dimensional wave-like pattern is generated on the skin. Image inspired by22,23,24. (c) The wave-like pattern of the acrylic adhesive at the bottom of a KT. (d) The one-dimensional wave-like pattern visible to the bare eye when a KT is applied to a leather sample used to mimic the skin. (e, f, g) Examples of KT applications on the knee (e) and on the calf (f). In (g) the application of a fan-shaped tape on the forearm for lymphatic drainage.
Hence, the elastic behavior of the tape, characterized by a close to zero Poisson’s ratio in combination with strong adhesion between the tape and the skin, creates an effect that recoils and folds up the skin in a longitudinal wave-like manner (Fig. 1b, d). However, since the weaving pattern of the fabric allows for the tape to expand only axially upon tension, the effect on the skin is one-dimensional, with no change in size in the transverse direction (Fig. 1d). Despite the inconclusive supporting evidence regarding the therapeutic efficacy of the KT7,13,25,26 for various treatments, it is still observed as an indispensable therapy amongst athletes and recreationists alike21.
The tape is usually applied by professionals or even patients themselves7,8. Due to its elastic behavior, in the majority of the cases, the tape is only applied along the length of the muscles, and the stimulus on the muscle is, as already underlined, one-dimensional27. Therefore, to treat larger affected areas, several pieces of KT are required (Fig. 1e, f, g)16. The application procedure thus comprises several manual steps involving cutting the KT, into multiple pieces, in combination with directional strategy applications to achieve the desired effect7,16. For example, to treat muscle tenseness in the forearm28 several pieces of tape are needed at the area of injury. In some instances, the pieces need to be cut to specific shapes making the application procedure complex, as in the case of fan-shaped strips to treat swellings (Fig. 1g)21. It is also worth underlying that the literature7 provides little or no evidence on how training and education of practitioners relate to treatment outcomes, especially in the case of more complex taping procedures.
Here, to simplify the application procedure of the tape to only one piece at the area of injury, instead of several, and to facilitate personalization, we propose a novel KT design based on auxetic structures. Auxetic structures are characterized by a negative Poisson’s ratio (v) which enables them to expand simultaneously in two perpendicular directions when a tension load is applied29,30. As such, auxetic structures found extensive applications in the development of textile-based smart wound dressings and bandages31,32. The structures are also well-known for their form-fitting properties, allowed by their capability to generate a synclastic curvature because of an out-of-plane bending29,33. The benefits of solutions capable of properly adhering to the body are privileged in the biomedical field. Indeed, for example, first 3D-printed fabrics reinforced with silicone are used to increase the efficacy of hypertrophic scar treatments34. Then, the beneficial effects of auxetic structures’ synclastic property, combined with a compressive bandage, are explored for the same objective35 in creating a stable level of pressure, during body motions, over the treated area34. When integrating an auxetic structure into a KT, the otherwise one-dimensional wave-like pattern created on the skin (Fig. 1b, d) is transformed into a two-dimensional one (Fig. 2). This can simplify the application procedure either for the physiotherapist or the end-user because it reduces the number of strips with complex directional strategies to be applied to achieve the positive decompressive effect (Fig. 1e, f, g), especially in cases of larger affected areas. The formfitting capability of the auxetic pattern is advantageous in terms of wearability and adhesion by reducing the risk of KT detaching because it allows the tape to properly adhere to curved body parts (e.g., joints), such as a knee or an elbow, both of which are parts prone to injures.
Novel auxetic KT (a) The positive decompressive effect generated by the auxetic KT application. A two-dimensional wave-like pattern is generated on the skin. Image inspired by22,23,24. (b) The two-dimensional wave-like pattern visible to the bare eye when a KT is applied on a leather sample used to mimic the skin.
A simplified procedure would lead to a more frequent application of the KT, which reduces the patient’s healing time: the simpler the application procedure is, the more motivated the patients are to apply the tape more often. Finally, the auxetic KT can be personalized for each patient by tuning the parameters of the auxetic design to change its mechanical behavior (i.e., the structure stiffness) and, consequently, the level of the decompressive effect it exerts on the injured area. The auxetic KT can be used to improve the treatment of the following injuries as it shows advantages over the taping methods with the existing KT: blood effusions, swellings, lymphatic blockages (in these three cases, the two-dimensional effect can be used to replace the complex fan-shaped taping, Fig. 1g); general pain at joints or muscles; pain points within muscles (the two-dimensional effect can replace the two tapes placed over the tension point); malpositioned muscles or joints (incorrect posture), fascia adhesion (the tape can cover a larger area and its effect is in multiple directions); scared tissue (with the auxetic KT only one piece is necessary instead of numerous to stimulate the skin). These additional benefits over the existing KT are likely to be well perceived by the physiotherapists and end-users of the tape as the market is rapidly growing, and there is a high demand for the product.
The reentrant honeycomb topology is selected to demonstrate the potential of the novel KT design. We started from this topology because in the literature its behavior has been extensively studied and formalized analytically with a focus on its main geometrical/design parameters. To prototype the design, first, an auxetic structure is fabricated by laser cutting an existing KT. Afterward, an additional auxetic structure is 3D-printed on the laser cut tape using thermoplastic polyurethane (TPU) material and fused filament fabrication (FFF) technology. In the first part of our work, analytical equations and experimental tests are used to determine the unit-cell geometry for the novel KT design based on the mechanical performance of five different reentrant honeycombs. In the second part, the behavior of an auxetic KT is experimentally compared to a standard KT. Prototypes are developed for two different types of injuries. One prototype is conceived to treat general elbow pain, while the other, muscle tenseness in the forearm.
When a tensional force acts on a reentrant honeycomb, three different deformation mechanisms operate on the unit cell: hinging (hg) at the cell nodes, flexing (f), and stretching (s) of the cell beams36,37,38. Combining these three effects, the Poisson’s ratio of the structure can be analytically calculated as [Eq. (1)]39,40,41:
where (theta), (l), (h), (b), (t) are geometrical parameters (Fig. 3) while (K_{{text{f}}}), (K_{{{text{hg}}}}), and (K_{{text{s}}}) [Eqs. (2)–(5)] are force constants that depend on the material properties (Young’s modulus (E_{{text{s}}}) and Poisson’s ratio (v_{{text{s}}}) of the material).
The reentrant honeycomb design and its main geometrical parameters.
The equations of the force constants are39,40,41:
where (q) is the axial length of the hinge (Fig. 3) and (G_{{text{s}}}) is the shear modulus, which can be derived from the Young’s modulus and the Poisson’s ratio of the material as [Eq. (6)]:
The hinging at the wall junctions, i.e., at the nodes, could occur for two mechanisms41: the shearing of the cell wall material [(K_{{{text{hg}}_{{{text{global}}}} }}), Eq. (2)] or the local bending of the wall [(K_{{{text{hg}}_{{{text{local}}}} }}), Eq. (3)].
The values of the geometrical parameters are assigned as ranges by also considering the technological feasibility in terms of the printability of the structure with the selected FFF machine and material (see the Methods Section). According to the reference formula used, the angle (theta) must be negative, (theta in left[ { – 60^{ circ } ,0^{ circ } } right]), the lengths of the diagonal (l) and vertical rib (h) must be in a defined range where (h,l in left[ {4{text{ mm}},12{text{ mm}}} right]), the thickness (t) should be a multiple of the printing nozzle diameter, (t = left[ {0.8{text{ mm}},1.2{text{ mm}},1.6{text{ mm}}, ldots } right]). Rib lengths should be at least five times as large as the rib thickness satisfying (5 – h/t le 0) and (5 – l/t le 0). The middle vertices of the unit cell must not overlap according to the following (1 – h – 2lsin theta + t/cos theta le 0). Finally, as in41, (q) can be set equal to (l/10).
The smallest Poisson’s ratio is found by graphically analyzing Fig. 4a–c containing the scatterplots of the Poisson’s ratio as a function of the geometrical parameters (theta), (h,) and (l) for fixed values of the parameters (t) and (b). They show that the smallest Poisson’s ratio is achieved by minimizing (theta), maximizing (h), and minimizing (l). By comparing the scatterplots for different values of (t), it can be noticed that a smaller (t) results in a more negative Poisson’s ratio; moreover, as (t) increases, the scatter plot contains fewer points as smaller ranges of (theta), (h,) and (l) fulfill the constraints. Figure 4d shows that the Poisson’s ratio remains approximately constant for a particular (h/l) ratio, while increasing in magnitude as the ratio becomes larger. Based on the previous considerations, five geometries of the auxetic structure are investigated (Table 1 and Fig. 5a) for the sensitivity of the Poisson’s ratio to each geometrical parameter. Tensile tests are conducted on 3D-printed five-by-five auxetic structures. The corresponding Poisson’s ratio of the structures is derived and calculated through image-based analysis (for further details, see the Methods Section). The values of the force constants and the Poisson’s ratio computed analytically are reported in Table 1, using the material properties of the TPU ((E_{{text{s}}}) = 26 MPa42, (v_{{text{s}}}) = 0.4843). In Fig. 5b, we report the Poisson’s ratio values of the five studied configurations as a function of the engineering strain. In this graph, the analytical values obtained, considering the (K_{{{text{hg}}_{{{text{global}}}} }}) force constant in Eq. (1), are also mapped since those are closer to the experimental values. These values do not depend on the engineering strain. Hence, they are represented as a line.
Scatterplots used to analyze the effect of the geometrical parameters (theta), (h,) and (l) on the Poisson’s ratio (nu), based on the analytical equation [Eq. (1), considering (K_{{{text{hg}}_{{{text{global}}}} }}) as hinging force constant], for fixed values of the (t) and (b) parameters, i.e., (t = 0.8 {text{mm}}) and (b = 1.5 {text{mm}}) (a), (t = 1.2 {text{mm}}) and (b = 1.5 {text{mm}}) (b), (t = 1.6 {text{mm}}) and (b = 1.5 {text{mm}}) (c), and for fixed values of (theta), (t) and (b) parameters (d, (theta = 20^circ ; t = 0.8 {text{mm}};b = 1.5 {text{mm}})).
Experimental results (a) The five different geometries of reentrant honeycombs investigated. (b) Experimental results of the Poisson’s ratio (nu) over the engineering strain (varepsilon) for the five geometries. In this chart also the Poisson’s value obtained from the analytical equations (global hinging) are mapped as lines (they do not depend on the engineering strain) (c) The four structures tested to measure the Poisson’s ratio, involving Structure 1 as a standard KT, Structure 2 realized as standard KT with auxetic perforations, Structure 3 as a 3D-printed auxetic structure, and Structure 4 as a combination of Structures 2 and 3 glued to each other. Structures 1 and 2 include v-shaped marks as relevant points for measuring the Poisson’s ratio (d) Results of the Poisson’s ratio (nu) over the engineering strain (varepsilon) for the four analyzed structures. In this chart, for Structure 3, also the Poisson’s value obtained from the analytical equations (global hinging) is mapped as a line.
The Poisson’s ratio of all geometries becomes less negative with increasing engineering strain. The comparisons between the curves show that decreasing (theta) (from Geometry 1 to Geometry 2) or decreasing (t) (from Geometry 3 to Geometry 4) results in a downward shift of the curve, increasing the magnitude of the Poisson’s ratio. Oppositely, by increasing the (h/l) ratio (from Geometry 2 to Geometry 3), the curve shifts upward, not matching the assumption from the analytical equation. This result occurs because by modifying the (h/l) ratio, (l) is decreased, but (t) is kept constant (Fig. 3). Hence, the (t/l) ratio increases, thus affecting the slenderness of the sloped walls of the cell which decreases. This change could affect the validity of the beam-based assumptions related to the analytical model. On the contrary, the slenderness of the vertical walls is kept constant because the (h) value does not change. When these ratios are preserved as in Geometry 1 and Geometry 2 there is almost a perfect match between the analytical and experimental values (Table 1). Indeed, Geometry 4 has an increase in the Poisson’s ratio compared to Geometry 3 because, in this case, the slenderness of both the vertical and sloped beams decreases (we have reduced (t), Table 1).
Theoretically, the prevalent deformation mechanism is the flexing (i.e., (K_{{text{f}}})) of the cell beams in all the studied geometries (i.e., it is the lowest K value which predominates41). However, the hinging at the nodes can become relevant when the length of the diagonal rib is reduced and (t) is kept constant. Besides, the change of the (l) value also affects the (q) value (as in41, it was set equal to (l/10)) in the (K_{{{text{hg}}_{{{text{local}}}} }}) [Eq. (3)]. It is also worth underlying that the shear modulus of the material in the hinge [Eq. (2)] is likely to be lower than the GS of the bulk material due to local damage41. Finally, the change of parameter b (from Geometry 3 to 5) does not influence the Poisson’s ratio. This result is confirmed both analytically and experimentally (see Table 1). Indeed, all force constants are directly proportional to b. Hence, the parameter b cancels out in Eq. (1).
The experimental results show that the Poisson’s ratio depends on the engineering strain and becomes less negative with increased strain. However, the analytical equation is independent of the engineering strain (they are represented as lines in Fig. 5b). The values obtained from the analytical equation represent the values from the experimental measurements at the lowest engineering strains (Table 1). Therefore, the analytical equation is only valid at small deformations. At larger deformations, the Poisson’s ratio is dependent on the level of strain, as already underlined by Wan et al.44. The error is more significant for Geometry 3, 4, and 5. This is most likely due to the ratio (l/t) not being sufficiently large. However, when comparing these three geometries to each other, the general assumptions from the analytical equation still hold.
Even though the analytical equation only provides a proper representation of the Poisson’s ratio at low engineering strains, it proves to be a good indicator, in general, to study whether specific geometrical parameters enlarge or reduce the auxetic effect. It is a simple procedure to determine how sensitive the auxetic effect is to the influence of the geometrical parameters.
From the results obtained from the experimental measurements (Fig. 5b), the following unit-cell dimensions are chosen: θ = -36°; h = 12 mm; l = 6.5 mm; tTPU = 0.8 mm; b = 1.5 mm; (t_{{{text{TAPE}}}} = 3.5) mm (Table 1). The unit-cell dimensions almost entirely equate the dimensions of Geometry 4, except for the length (l) which is slightly increased to be able to design a perforated KT such that it covers a large enough area of patient’s skin to secure a proper adhesion. The value 6.5 mm is derived from applying the constraint which ensures that the middle corners of the unit cell do not overlap with the tape. The selected geometry both maximizes the auxetic effect and, at the same time, controls the out-of-plane deformation upon stretching of the overall auxetic structure.
A second experimental campaign is then conducted to evaluate the differences in Poisson’s ratios of the standard and the auxetic KT based on the selected unit-cell geometry. Thus, four structures are investigated (Fig. 5c). The first consists of the standard KT. The second is fabricated by laser cutting a standard KT to create an auxetic perforation. In this case, we select (t_{{{text{TAPE}}}} = 3.5) mm where (t_{{{text{TAPE}}}}) corresponds to the rib thickness (t) value of the unit-cell geometry. The (t_{{{text{TAPE}}}} = 3.5) mm value is set higher than (t_{{{text{TPU}}}}) so that the tape covers a larger area of the patient’s skin for proper adhesion. The third structure (Fig. 5c) is a 3D-printed auxetic structure, while the fourth structure is the combination of the second and third structures comprising a laser-cut KT with a TPU auxetic structure 3D-printed on top of it (Fig. 5c). Figure 5d reports the results of tensile tests showing Poisson’s ratio over the engineering strain of these four structures. Structure 1 exhibits a Poisson’s ratio, which is close to 0 and independent of the strain, in accordance with the literature and manufacturers’ indications related to standard KT. Structure 2 has a positive Poisson’s ratio, meaning that the laser-cut KT by itself does not show any auxetic behavior. Structure 4 shows a Poisson’s ratio that is less negative than Structure 3, as the tape realized as Structure 2 does not contribute to the auxetic behavior, in fact it does the opposite. The Poisson’s value obtained from the analytical equations (global hinging) is mapped as a line for Structure 3. As explained in the Methods Section, in the case of Structure 4 some values are not provided in this chart because non-planar deformations occurred at those points.
Although the application of KT covers many types of injuries, among them, two different but frequent application scenarios are chosen to explore the potential of the novel auxetic KT. In the first case, the auxetic KT is designed to treat general pain in the elbow due to overuse or overstretching of the elbow. Often, the patient experiences dull pain, which is not explicitly located in one area and can result in weakness of the joint. Therefore, a taping method for general elbow pain, which provides support and relieves pain by lifting the skin so that the patient can continue with activities, proves helpful. When using the traditional KT (Fig. 6a), the elbow is flexed, and one end of the first piece of tape is attached below the elbow with no tension. It is then applied along the medial side of the elbow with a stretch of 50% of the maximum possible tension of the tape. The second end is attached above the elbow with no stretch. The same procedure is repeated with another piece of tape along the lateral side of the elbow. When an auxetic structure is incorporated into the KT, the synclastic curvature will allow the tape to conform to the shape of the elbow (Fig. 6b). This effect permits the use of only one piece of auxetic tape instead of several parts. This leads to a simplification of the application procedure.
Comparison between the application procedure of the traditional KT and the novel auxetic KT for two cases. (a) The traditional KT application procedure for general elbow pain divided into 4 steps. (b) The auxetic KT for elbow pain. (c) The traditional KT application procedure for forearm muscle tenseness divided into 4 steps. (d) The auxetic KT for forearm muscle tenseness.
The second case is the application of the auxetic KT on the muscles of the forearm to treat muscle tenseness. This pain is caused by overuse or inflammation of the muscle, which creates a point of tension within the muscle. With the traditional KT (Fig. 6c), the muscle should be stretched by flexing the wrist into a downward position21. The first end of the tape is applied to the hand with zero tension. The tape is then applied along the length of the arm with a stretch of 25–50% of the maximum possible tension. No tension is applied on the second end of the tape. The stretching of the tape causes a convolution in the skin to reduce the pressure and relieve the pain. At the location of pain, a second shorter piece of tape is applied perpendicular to the first piece of tape, which aids in lifting the skin even more (Fig. 6c, Step 4). Using an auxetic structure at the intersection allows combining two pieces of tape into one; therefore, it simplifies the application procedure (Fig. 6d).
Based on these two applications, the following considerations were derived. In the case of the general elbow pain (Fig. 6a, b), the location of the auxetic KT compared to the standard KT is different. With the existing KT, the two pieces are placed around the elbow joint (Fig. 6a). Instead, the auxetic KT is placed directly over the joint (Fig. 6b). In this case, the joint can move freely, as the property of synclastic curvature allows a good adaptation of the tape to the shape of the joint and ensures unrestricted movements. An end-user could intuitively think that the auxetic tape restricts movements, as it covers the entire joint. However, when the tape is applied instead, the elbow can be moved freely (Fig. 6b). This proves that the synclastic curvature of the auxetic tape adds benefits to a KT.
Besides lifting the skin, the auxetic tape can also benefit the fascia (Fig. 1a, b). The fascia is the connective tissue underneath the skin, which encloses the muscles. Compared to the muscles, the fascia can cover a larger area. A standard KT, in most cases, is applied along the length of the muscles, and the stimulus on the muscle is one-dimensional. However, the auxetic KT can cover the larger area of the fascia. Therefore, the expansion of the auxetic tape in two directions is also valid for treating the fascia. The auxetic KT can thus support the fascia as it increases its tension. Furthermore, errors within the application procedure are reduced when using the auxetic KT to treat the fascia. These errors can occur when larger areas, such as the shoulder, are treated. Existing KT cannot cover such larger areas. Several longitudinal tapes would be needed, but the amount of tension in each tape should be the same. If this is not the case, one tape takes away the function of the other, leading to incorrect treatment. This issue can be avoided when using only one auxetic tape to cover the whole area.
Regarding the fabrication of the auxetic KT, an easy application requires that similarly to the existing KT, it is fabricated as one long strip and delivered on a roll. A physiotherapist or an athlete could thus easily cut the required length of tape from the roll. The current auxetic prototypes were fabricated using a 3D printing technology that does not allow the manufacturing of such rolls. Further fabrication techniques could be explored, such as the manufacturing of auxetic fabrics or auxetic yarn45,46. On one side of the auxetic fabric, an acrylic adhesive should be applied to guarantee the proper attachment to the skin. Besides, the rigidity of the auxetic tape should be adequately tuned, by adjusting the structure stiffness or creating auxetic KT from auxetic fabric, woven or 3D-printed. However, the auxetic KT rigidity represents an advantage for its personalization. For example, an increased rigidity might be helpful when an athlete is in the acute phase of the injury, during which no sports activities are allowed.
Therefore, the tape mixes the original inelastic athletic tape, which restricts movements, and the elastic KT, which allows free movements. The athlete can freely move due to the synclastic curvature, but still has increased injury support thanks to the auxetic structure. By properly designing this structure, the novel tape could also provide different levels of treatment.
A novel KT has been developed based on auxetic structures. A combination of laser cutting and 3D printing is used to fabricate the novel auxetic tape. The Poisson’s ratio is investigated through analytical equations and experimental tests. The auxetic KT proves to reduce the steps of the application procedure because it generates a two-dimensional treatment, provides comfort by allowing a free range of motion due to the synclastic property of auxetics, allows tuning the tape rigidity to allow personalizing its effect to the end-user’s skin and type of injury.
The novel KT has been designed using the reentrant honeycomb topology as a reference. Due to its simplicity, it allowed to use FFF to 3D print the structure and to laser cut the shape into the tape easily. However, further topologies can be also explored, involving auxetic structures with chiral building-blocks47, superelliptic parameterizations48, or even auxetic fabrics31, and as such creating geometries with fewer sharp corners. In that way, the selection of the topology could also consider heterogenous design aspects such as comfort but also fabrication constraints.
The novel auxetic KT is composed of two parts; an auxetic laser-cut KT and a corresponding 3D-printed TPU auxetic structure.
First, the auxetic structure is fabricated by laser cutting (Model: Adcomm VREL Laser CO2-N 600/80, Software: RDWorksV8) a standard KT (Fig. 7a). Then, an auxetic structure built using the flexible TPU filament (Ultimaker TPU 95A42) is 3D-printed directly onto the tape (Fig. 7b, c) using FFF technology (Ultimaker S5, printing nozzle 0.4 mm diameter). The auxetic structures for the laser cutting and those for the 3D printing are modeled using Grasshopper 3D, which runs within the computer-aided design software Rhinoceros 3D (Version 7.4).
Fabrication steps of the auxetic KT. (a) The auxetic structure is laser cut into the KT. (b) The flexible material (TPU) is 3D-printed onto the tape. (c) Fabricated auxetic KT prototype. (d) Supporting rectangles and horizontal bars are added to the auxetic structures used for the mechanical characterization. (e) v-shaped marks indicated on the only tape samples (Structure 1 in Fig. 5c) as relevant points for measuring the Poisson’s ratio. (f) The sample used for Structure 2 (Fig. 5c) and a detail of the v-shaped marks.
To prepare the 3D printing pre-processing files, the Ultimaker Cura slicer software (Version 4.8.0) is used. The standard printing settings for the Ultimaker TPU 95A filament are used with a few exceptions; the infill density of the filament is changed to 100%, whereas the z-offset is set to 0.5 mm, to roughly approximate the thickness of the tape. The tape needs to be placed precisely at the position on the printing bed, where the TPU filament will be printed, according to the Ultimaker Cura software. For all the four auxetic structures analyzed (Fig. 5c), supporting rectangles (Fig. 7d) are included in the design to position them in the tensile testing device and provide proper clamping during measurements. Along the edge of the bottom rectangle, three horizontal bars are added, ensuring that the structure is positioned straight (Fig. 7d). When comparing the existing and auxetic KT, v-shaped marks were drawn on Structure 1 and Structure 2 in Fig. 5c using a laser-cut mask (Fig. 7e, f) to properly define the points used for the Poisson’s ratio calculation through image-based analysis. Figure 7d also shows the type of samples that were 3D-printed and tested to obtain the Poisson’s ratio (Fig. 5b) of the five reentrant honeycomb topologies shown in Fig. 5a. In this case, the dimensions of the supporting rectangles vary with the auxetic topology.
For the tuning of the auxetic structure design, tensile tests (Electromechanical Testing Systems, MTS Synergie 200, load cell 1kN) are performed on an auxetic structure containing five unit cells in both x– and y-direction. The test setup is shown in Fig. 8a. While the measurement device applies a tensile load with a speed of 5 mm/min on the structure, pictures are taken at different engineering strains with the Sony A6000 camera. The supporting rectangles of the structures (Fig. 7d) are clamped to the measurement device. The clamping of these parts restricts the auxetic behavior of the structure as it cannot expand in this area. Therefore, the Poisson’s ratio is determined for the central unit cell of the structure (highlighted in Fig. 8b), which is influenced the least by this end effect. Each picture (resolution of 350 × 350 dpi, containing 4000 × 6000 pixels), taken at specific engineering strains, is compared to the original undeformed image to calculate the Poisson’s ratio of the unit cell. There is no need to stop the testing machine at different strains to take pictures. The experiment runs continuously until deformation of 20 mm is reached. The camera is programmed to take a photo every 9 s for 30 shots in total. These images are analyzed in Matlab (Version Matlab R2020b). Four relevant points (labeled simply using numbers from 1 to 4 in Fig. 8b) are determined in the middle of the vertical ribs of the unit cell. Since they cannot be marked directly by hand, the corner points (Ai and Bi, i = 1, 2, 3, 4 shown in Fig. 8b) are marked, and their coordinates are calculated in pixels. Hence, the coordinates of points labeled using numbers from 1 to 4 in Fig. 8b can be calculated. From these four points, the distances (X_{0}) and (Y_{0}) can be computed for the undeformed structure, and (X_{{text{D}}}) and (Y_{{text{D}}}) for the deformed one (Fig. 8c). Finally, the Poisson’s ratio is determined as (nu = – frac{{varepsilon_{x} }}{{varepsilon_{y} }} = – frac{{left( {X_{{text{D}}} – X_{0} } right)/X_{0} }}{{left( {Y_{{text{D}}} – Y_{0} } right)/Y_{0} }}). This calculation is done for the image taken at each strain (calculated using the (L_{0}) and (L_{{text{D}}}) measurements shown in Fig. 8c as (varepsilon = frac{{L_{{text{D}}} – L_{0} }}{{L_{0} }})) to plot the Poisson’s ratio over the engineering strain. For each auxetic structure, three samples are tested. The average Poisson’s ratio is calculated for the central unit cell of the structure and plotted over the engineering strain. In addition, at each strain, the standard deviation of the three samples is calculated to evaluate the variance and consequently the repeatability.
Mechanical characterization and data analysis. (a) Tensile test setup for the mechanical characterization. (b) Data analysis: the Poisson’s ratio is determined for the highlighted central unit cell. The coordinates of relevant points (labeled as Ai and Bi, i = 1, 2, 3, 4) are computed in pixels, from which the coordinates of the middle point of each vertical beam (labeled simply using numbers from 1 to 4) are calculated. (c) Undeformed and deformed auxetic structure in the tensile test measurement device: (X_{0}), (Y_{0}), (X_{{text{D}}}) and (Y_{{text{D}}}) are used to calculate the Poisson’s ratio at the given strain, determined using the (L_{0}) and (L_{{text{D}}}) measurements. (d) Undeformed and deformed auxetic KT. The points analyzed for this second experiment are also highlighted.
For the second experiment related to the comparison between the standard and the auxetic KT, we design a five-by-seven structure. We use this higher number of cells to have a longer tape and thus reduce the influence of the clamping, especially in the case of the tape-only structure (i.e., Structure 1 in Fig. 5c). In this case, we do not analyze the Poisson’s ratio of the central unit cell, but the Poisson’s ratio of a larger area (Fig. 8d). This choice is made because this area can represent the possible extent of the patient’s injury whose Poisson’s ratio is what matters for the application. Besides, it is also small enough to decrease the influence of the clamping.
Figure 8d, right, also shows the reason why fewer points are plotted in the chart in Fig. 5d for Structure 4. At a certain point a non-planar deformation occurs, and the calculation of the Poisson’s value is no longer valid. The five-by-seven structures are tested at 7 mm/min until a deformation of 28 mm is reached. The setup for the camera is the same as in the previous experiment.
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.
Nippert, A. H. & Smith, A. M. Psychologic stress related to injury and impact on sport performance. Phys. Med. Rehabil. Clin. N. Am. 19, 399–418 (2008).
PubMed  Article  Google Scholar 
Hootman, J. M. et al. Epidemiology of musculoskeletal injuries among sedentary and physically active adults. Med. Sci. Sports Exerc. 34, 838–844 (2002).
PubMed  Article  Google Scholar 
Smith, A. M. & Milliner, E. K. Injured athletes and the risk of suicide. J. Athl. Train. 29, 337–341 (1994).
CAS  PubMed  PubMed Central  Google Scholar 
Sonstroem, R. J. & Morgan, W. P. Exercise and self-esteem: Rationale and model. Med. Sci. Sports Exerc. 21, 329–337 (1989).
CAS  PubMed  Article  Google Scholar 
Brewer, B. W., Van Raalte, J. L. & Linder, D. E. Athletic identity: Hercules’ muscles or achilles heel?. Int. J. Sport Psychol. 24, 237–254 (1993).
Google Scholar 
Brewer, B. W. Self-identity and specific vulnerability to depressed mood. J. Pers. 61, 343–364 (1993).
CAS  PubMed  Article  Google Scholar 
Cheatham, S. W., Baker, R. T. & Abdenour, T. E. Kinesiology tape: A descriptive survey of healthcare professionals in the united states. Int. J. Sports Phys. Ther. 16, 778–796 (2021).
PubMed  PubMed Central  Google Scholar 
Nelson, N. L. Kinesio taping for chronic low back pain: A systematic review. J. Bodyw. Mov. Ther. 20, 672–681 (2016).
PubMed  Article  Google Scholar 
Kase, K. Illustrated Kinesio Taping Manual. (Ken Ikai Co Ltd, 2005).
Skirven, T. M., Osterman, A. L., Fedorczyk, J. M. & Amadio, P. C. Rehabilitation of the Hand and Upper Extremity. (Mosby Inc., an affiliate of Elsevier Inc., 2011).
Wu, W. T., Hong, C. Z. & Chou, L. W. The kinesio taping method for myofascial pain control. Evidence-based Complement. Altern. Med. 2015, (2015).
Ristow, O. et al. Does elastic therapeutic tape reduce postoperative swelling, pain, and trismus after open reduction and internal fixation of mandibular fractures?. J. Oral Maxillofac. Surg. 71, 1387–1396 (2013).
PubMed  Article  Google Scholar 
González-Iglesias, J., Fernández-De-Las-Peñas, C., Cleland, J., Huijbregts, P. & Del Rosario-Gutiérrez-Vega, M. Short-term effects of cervical kinesio taping on pain and cervical range of motion in patients with acute whiplash injury: A randomized clinical trial. J. Orthop. Sports Phys. Ther. 39, 515–521 (2009).
PubMed  Article  Google Scholar 
Montalvo, A. M., Cara, E. L. & Myer, G. D. Effect of kinesiology taping on pain in individuals with musculoskeletal injuries: Systematic review and meta-analysis. Phys. Sportsmed. 42, 48–57 (2014).
PubMed  Article  Google Scholar 
Mostafavifar, M., Wertz, J. & Borchers, J. A systematic review of the effectiveness of kinesio taping for musculoskeletal injury. Phys. Sportsmed. 40, 33–40 (2012).
PubMed  Article  Google Scholar 
Choi, I. R. & Lee, J. H. Effect of kinesiology tape application direction on quadriceps strength. Med. 97, e11038 (2018).
Article  Google Scholar 
Konishi, Y. Tactile stimulation with Kinesiology tape alleviates muscle weakness attributable to attenuation of Ia afferents. J. Sci. Med. Sport 16, 45–48 (2013).
PubMed  Article  Google Scholar 
Hosp, S. et al. A pilot study of the effect of Kinesiology tape on knee proprioception after physical activity in healthy women. J. Sci. Med. Sport 18, 709–713 (2015).
PubMed  Article  Google Scholar 
Williams, S., Whatman, C., Hume, P. A. & Sheerin, K. Kinesio Taping in Treatment and Prevention of Sports Injuries. Sport. Med. 42, 153–164 (2012).
Article  Google Scholar 
Kumbrink, B. K-Taping. (Springer, 2014). https://doi.org/10.1007/978-3-662-43573-1.
Kase, K., Wallis, J. & Kase, T. Clinical therapeutic application of the kinesio taping method. Ken Ikai Co Ltd (2013).
Kase, K. & Stockheimer, K. R. Kinesio Taping for Lymphoedema and Chronic Swelling Manual. (Kinesio Taping).
CV Integrative therapies – Kinesiology Taping. http://www.cvintegrativetherapies.com/integrative-therapy.html.
Gibbons. J. Practical Guide to Kinesiology Taping for Injury Prevention and Common Medical Conditions. (Lotus Pub, 2019).
Bac, A., Wróbel, M., Ogrodzka-Ciechanowicz, K., Michalik, E. & Ścisłowska-Czarnecka, A. The influence of the Kinesio Taping on selected ultrasonography measurements, and quality of life in patients with rotator cuff lesions. Sci. Rep. 10, 1–8 (2020).
Article  Google Scholar 
Noguera-Iturbe, Y., Martínez-Gramage, J., Montañez-Aguilera, F. J., Casaña, J. & Lisón, J. F. Short-term effects of kinesio taping in the treatment of latent and active upper trapezius trigger points: Two prospective, randomized, Sham-Controlled Trials. Sci. Rep. 9, 1–7 (2019).
CAS  Article  Google Scholar 
Liu, Y. & Hu, H. A review on auxetic structures and polymeric materials. Sci. Res. Essays 5, 1052–1063 (2010).
Google Scholar 
Rocktape instruction. https://www.rocktape.com/wp-content/uploads/2018/12/ENGLISH_ROCKTAPE_RockTape_instructios.pdf.
Lim, T.-C. Auxetic Materials and Structures. (Springer Singapore, 2015). https://doi.org/10.1007/978-981-287-275-3.
Lim, T.-C. Mechanics of Metamaterials with Negative Parameters. (Springer Singapore, 2020). https://doi.org/10.1007/978-981-15-6446-8.
Hu, H., Zhang, M. & Liu, Y. Auxetic textiles. Auxetic Textiles (Woodhead Publishing, 2019). https://doi.org/10.1016/C2016-0-04399-1.
Wang, Z. & Hu, H. Auxetic materials and their potential applications in textiles. Text. Res. J. 84, 1600–1611 (2014).
Article  Google Scholar 
Evans, K. E. & Alderson, A. Auxetic materials: Functional materials and structures from lateral thinking!. Adv. Mater. 12, 617–628 (2000).
CAS  Article  Google Scholar 
Chow, L. et al. Customized fabrication approach for hypertrophic scar treatment: 3D printed fabric silicone composite. Int. J. Bioprinting 6, 1–12 (2020).
Article  Google Scholar 
Chow, L. et al. 3D printing auxetic architectures for hypertrophic scar therapy. Macromol. Mater. Eng. 307, 1–12 (2022).
Article  Google Scholar 
Gibson, L. J. & Ashby, M. F. The mechanics of honeycombs. in Cellular Solids 93–174 (Cambridge University Press, 2014). https://doi.org/10.1017/cbo9781139878326.006.
Grima, J. N., Cauchi, R., Gatt, R. & Attard, D. Honeycomb composites with auxetic out-of-plane characteristics. Compos. Struct. 106, 150–159 (2013).
Article  Google Scholar 
Mitschke, H. et al. Geometry: The leading parameter for the Poisson’s ratio of bending-dominated cellular solids. Int. J. Solids Struct. 100–101, 1–10 (2016).
Article  Google Scholar 
Whitty, J. P. M., Nazare, F. & Alderson, A. Modelling the effects of density variations on the in-plane Poisson’s ratios and Young’s moduli of periodic conventional and re-entrant honeycombs – Part 1: Rib thickness variations. Cell. Polym. 21, 69–98 (2002).
Google Scholar 
Evans, K. E., Alderson, A. & Christian, F. R. Auxetic two-dimensional polymer networks: An example of tailoring geometry for specific mechanical properties. Faraday Trans. 91, 2671–2680 (1995).
CAS  Article  Google Scholar 
Masters, I. G. & Evans, K. E. Models for the elastic deformation of honeycombs. Compos. Struct. 35, 403–422 (1996).
Article  Google Scholar 
Ultimaker BV. Ultimaker TPU 95A tds. https://support.ultimaker.com/hc/en-us/articles/360012664440-Ultimaker-TPU-95A-TDS.
Qi, H. J. & Boyce, M. C. Stress-strain behavior of thermoplastic polyurethanes. Mech. Mater. 37, 817–839 (2005).
Article  Google Scholar 
Wan, H., Ohtaki, H., Kotosaka, S. & Hu, G. A study of negative Poisson’s ratios in auxetic honeycombs based on a large deflection model. Eur. J. Mech. A/Solids 23, 95–106 (2004).
MATH  Article  Google Scholar 
Sloan, M. R., Wright, J. R. & Evans, K. E. The helical auxetic yarn – A novel structure for composites and textiles; Geometry, manufacture and mechanical properties. Mech. Mater. 43, 476–486 (2011).
Article  Google Scholar 
Zhao, S. et al. Deformation behaviors of auxetic warp knitted fabrics based on reentrant geometry. Phys. Status Solidi Basic Res. 2000580, 1–11 (2021).
Google Scholar 
Grima, J. N., Gatt, R. & Farrugia, P. S. On the properties of auxetic meta-tetrachiral structures. Phys. Status Solidi Basic Res. 245, 511–520 (2008).
ADS  CAS  Article  Google Scholar 
Clausen, A., Wang, F., Jensen, J. S., Sigmund, O. & Lewis, J. A. Topology optimized architectures with programmable Poisson’s ratio over large deformations. Adv. Mater. 27, 5523–5527 (2015).
CAS  PubMed  Article  Google Scholar 
Download references
Engineering Design and Computing Laboratory, Department of Mechanical and Process Engineering, ETH Zürich, Tannenstrasse 3, 8092, Zürich, Switzerland
Luna Meeusen & Tino Stanković
Department of Mechanical Engineering, Politecnico Di Milano, Via La Masa 1, 20156, Milan, Italy
Sara Candidori, Laura Loredana Micoli & Serena Graziosi
Luddy School of Informatics, Computing and Engineering, Luddy Center for Artificial Intelligence, Indiana University, 1015 E 11th St, Bloomington, 47408, USA
Gabriele Guidi
You can also search for this author in PubMed Google Scholar
You can also search for this author in PubMed Google Scholar
You can also search for this author in PubMed Google Scholar
You can also search for this author in PubMed Google Scholar
You can also search for this author in PubMed Google Scholar
You can also search for this author in PubMed Google Scholar
L.M. designed the new tape. L.M. and S.C. contributed to the prototyping. All authors contributed to the experiments. L.M, S.C., T.S. and S.G. analysed the results. T.S. and S.G. supervised the project. All authors reviewed and approved the manuscript.
Correspondence to Serena Graziosi.
The authors declare no competing interests.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Reprints and Permissions
Meeusen, L., Candidori, S., Micoli, L.L. et al. Auxetic structures used in kinesiology tapes can improve form-fitting and personalization. Sci Rep 12, 13509 (2022). https://doi.org/10.1038/s41598-022-17688-w
Download citation
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41598-022-17688-w
Anyone you share the following link with will be able to read this content:
Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.
Advertisement
Advanced search
Scientific Reports (Sci Rep) ISSN 2045-2322 (online)
© 2022 Springer Nature Limited
Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

source

Issa Rae Headlines American Express’ Business Class LIVE 2022 In New York City – Forbes

Issa Rae at The 2022 Business Class Live with American Express
Fear of a possible recession and rising inflation has small business owners thinking how they might make prospective pivots sooner than later.
For its historic tenth installment of the program, American Express, a steadfast supporter of small businesses, is bringing back the famed Business Class LIVE conference. On July 20, a full-day event was conducted in the Javits Center in New York, and attendance in person or online was free of charge. This free event was created to inspire and support small business owners and leaders from all walks of life with strong connections and expert insights. It brings the energy and excitement of in-person panels and workshops with experts on cash flow management, marketing strategies, contracting opportunities, and workplace wellness for small business owners from all over the country..
The Business Class LIVE event connected attendees with other business owners, helping them learn from industry experts and entrepreneurs, and giving them the edge needed to face the road ahead with confidence.
A star-studded lineup of speakers and panelists, many of whom were minority and female owned business leaders and small business owners. Among which were:

Clayton Ruebensaal, executive vice president, global B2B marketing at American Express with Insecure … [+] creator and star Issa Rae at the 2022 American Express Business Class Live
For(bes) The Culture had the chance to speak with Clayton Ruebensaal, executive vice president, global B2B marketing at American Express, about what the conference comprised in addition to famed Insecure creator and star Issa Rae.
Ruebensaal spoke about how the company is committed to supporting small business and discussed the state of small businesses, especially with supply chain issues, a high level of open employment opportunities and the possibility of a recession.
According to Ruebensaal Amex aims to help businesses do more business. “To address the most important issues facing small business owners, we created Business Class. We were aware that when the pandemic struck, we had to step up our efforts since small businesses were struggling with previously unexplored problems and lacked information. The conference, which wass free to attend, is a ground-breaking, international initiative that redefines what business education is by providing small business owners with experiences and content that are more in line with their lifestyles. Its goal is to democratize access to the information that is most essential.”
Issa Rae, actress, producer, and entrepreneur at the 2022 American Express Business Class Live
Rae added her thoughts on the significance of the Business Class LIVE and expressed her excitement at the opportunity to represent small business owners.
“Just being able to look people in the eye is exciting to me. There is nothing more fulfilling than interacting with others and seeing their faces. I place a lot of emphasis on leveraging the individuals in your immediate surroundings. Being small business owners, we can all learn from one another and provide one another with advice because in a sense, we’re all in the same boat. I myself run a small business. It was an honor to obtain knowledge from the interactions I engaged in with other minority business owners,” expressed Rae.
Ruebensaal, adds the event offers a fantastic opportunity to forge long-lasting connections, learn new skills, and build a reliable network that can be used long after the event.
“It will allow small businesses the chance to meet one-on-one with representatives from governmental organizations or large corporations to talk about actual, concrete business opportunities that could help them in significant ways.”
Rae is a strong advocate for making it easy for consumers to find Black-owned businesses, but she also discusses how women-owned and minority-led businesses should support one another.
“You cannot be threatened by other businesses. We at Hilltop are located close to another Black-owned coffee shop. It costs us nothing to promote that other shop and tell customers to go there as is; in return, they will do the same for us. My show Sweet Life is shot in both Sip & Sonder and Hilltop, in addition to Haron coffee. Supporting one another as Black businesses will only benefit our businesses. While it’s good that we sometimes have a competitive attitude, it’s also crucial that we support one another,” tells Rae.
While Rae is well recognized for supporting everything and everyone Black, the objective is to get other larger corporations besides American Express on trend to move the needle forward.
Rae tells us American Express has been very intentional over the last couple of years, highlighting Black businesses.
“For many years, they have supported small businesses. I believe that other businesses should take a cue from Amex and make this a year-round initiative rather than just something that happens during Black History Month. Amex has been very deliberate about keeping in touch with the business owners I recommend to them. It is crucial and might mean the difference between being able to feed your family or not to have a chance and a platform to attract fresh eyes and customers to their services or business,” expresses Rae.
Rae plans to help expand her small business, Hilltop Coffee to LAX. The coffee shop currently has 3 locations in Los Angeles, Inglewood, and Eagle Rock.
“I’m overjoyed right now. We’ll be at LAX as the go-to place whenever you’re hungry, need a coffee or a snack. Being able to represent LA at the airport as a local business is amazing. We’ll be close to Randy’s Donuts, another staple in Inglewood. It’s very satisfying to know travelers will recognize that the restaurant they’re purchasing from is a LA Black-owned establishment. I’m appreciative of this chance. It’s not always the case that we have access to these stores in that way. There was intention behind this move from the city of LA. We feel honored to be there,” explains Rae.

source

Prime Video Taps Michael Smith as News Analyst For Thursday Night Football Coverage – Amazon Studios


Distinguished journalist, reporter, and host to provide insights and latest news from around the NFL throughout Thursday Night Football’s wraparound coverage
Prime Video’s inaugural Thursday Night Football season kicks off on September 15 with Chargers at Chiefs
NEW YORK—August 9—Prime Video’s Thursday Night Football continues to bolster its on-air talent roster for the upcoming season, as the streamer announced today that Michael Smith will join its coverage as its Thursday Night Football news analyst.

In this role, Smith will deliver and analyze the latest news from around the NFL, and advance pertinent stories with insightful commentary and information throughout Thursday Night Football’s pregame, halftime, and postgame coverage.

“Michael Smith is one of sports’ most respected reporters, and we look forward to him bringing his storytelling and insights to our Thursday Night Football coverage,” said Marie Donoghue, vice president, Global Sports Video, Amazon.
“I’ve spent more than two decades covering dynasties and super teams—now I finally get to be a part of one,” said Smith. “The talent and experience Prime Video have assembled on-camera, behind the camera, and throughout the entire TNF team qualifies as a bona fide Dream Team, and it’s an actual dream come true for me to be a part of a historic, groundbreaking prime-time NFL broadcast.”

Michael Smith is one of sports media’s most respected, versatile, and accomplished storytellers. Since 2020, he’s been a host and commentator for NBC Sports, serving as co-host and executive producer of the sports, current events, and culture show Brother From Another on Peacock, and contributed to NBC’s pregame coverage of Super Bowl LVI. Previously, over 15 years at ESPN, Michael served as a reporter, commentator, host, and anchor. He began his time in Bristol as an NFL insider and E:60 correspondent, before going on to co-host his own daily talk show, His & Hers, and co-anchoring SportsCenter. Prior to joining ESPN, Michael was a reporter for The Boston Globe, serving as the paper’s NFL and New England Patriots beat writer. He is also the co-founder of the California- and Connecticut-based content studio, Inflection Point Entertainment, and is a member of the 2017 Class of Henry Crown Fellows within the Aspen Global Leadership Network. Smith began his career while a student at Loyola University in his hometown of New Orleans. He received his B.S. in mass communication from Loyola in 2001.

Smith joins Thursday Night Football’s all-star broadcast team that includes legendary play-by-play announcer Al Michaels, five-time Emmy-winning analyst Kirk Herbstreit, award-winning reporter Kaylee Hartung, Pro Football Hall of Famer Tony Gonzalez, All-Pro cornerback Richard Sherman, NFL quarterback extraordinaire Ryan Fitzpatrick, NFL expert and top sports television host Charissa Thompson, contributors and Super Bowl Champions Andrew Whitworth and Aqib Talib, and emerging star Taylor Rooks as feature reporter. Fred Gaudelli, producer of seven Super Bowls and a recent inductee into the Sports Broadcasting Hall of Fame, will serve as the executive producer of Thursday Night Football’s game coverage. Multi-Emmy-award-winners Mike Muriano and Spoon Daftary are the executive and senior coordinating producers of Thursday Night Football’s pregame, halftime, and postgame coverage.
Kicking off September 15, Prime Video will become the first streaming service to air a season-long exclusive national broadcast package with the NFL. The 11-year deal includes 15 regular-season games and one preseason game per year, with Prime Video also delivering new pregame, halftime, and postgame shows as well as fan-favorite interactive features like X-Ray and Next Gen Stats powered by AWS. Viewers can stream from the web at amazon.com or by using the Prime Video app, which is available on smartphones, tablets, set-top boxes, game consoles, and connected TVs. Each Thursday Night Football game will also stream live on Twitch and will be available on over-the-air TV stations in the two home markets of the competing teams. For all the latest information regarding TNF on Prime Video, visit amazon.com/TNF and follow NFL on Prime on Twitter. To join Prime or start a free 30-day trial, visit amazon.com/prime.

Thursday Night Football Schedule:
Pregame Coverage Begins at 7 p.m. EDT. Kickoff Time 8:15 p.m. EDT.

Week 2 – Sept. 15: Los Angeles Chargers at Kansas City Chiefs (GEHA Field at Arrowhead Stadium, Kansas City, MO)
Week 3 – Sept. 22:
Pittsburgh Steelers at Cleveland Browns (FirstEnergy Stadium, Cleveland, OH)
Week 4 – Sept. 29:
Miami Dolphins at Cincinnati Bengals (Paul Brown Stadium, Cincinnati, OH)
Week 5 – Oct. 6:
Indianapolis Colts at Denver Broncos (Empower Field at Mile High, Denver, CO)
Week 6 – Oct. 13:
Washington Commanders at Chicago Bears (Soldier Field, Chicago, IL)
Week 7 – Oct. 20:
New Orleans Saints at Arizona Cardinals (State Farm Stadium, Glendale, AZ)
Week 8 – Oct. 27:
Baltimore Ravens at Tampa Bay Buccaneers (Raymond James Stadium, Tampa, FL)
Week 9 – Nov. 3:
Philadelphia Eagles at Houston Texans (NRG Stadium, Houston, TX)
Week 10 – Nov. 10:
Atlanta Falcons at Carolina Panthers (Bank of America Stadium, Charlotte, NC)
Week 11 – Nov. 17:
Tennessee Titans at Green Bay Packers (Lambeau Field, Green Bay, WI)
Week 13 – Dec. 1:
Buffalo Bills at New England Patriots (Gillette Stadium, Foxborough, MA)
Week 14 – Dec. 8:
Las Vegas Raiders at Los Angeles Rams (SoFi Stadium, Inglewood, CA)
Week 15 – Dec. 15:
San Francisco 49ers at Seattle Seahawks (Lumen Field, Seattle, WA)
Week 16 – Dec. 22:
Jacksonville Jaguars at New York Jets (MetLife Stadium, East Rutherford, NJ)
Week 17 – Dec. 29:
Dallas Cowboys at Tennessee Titans (Nissan Stadium, Nashville, TN)
About Prime Video Sports
Prime Video supports a growing lineup of live sports globally, including Thursday Night Football, the WNBA, the Seattle Storm, the New York Yankees, and Seattle Sounders FC in the United States; Premier League in the United Kingdom; US Open Tennis, ATP, and WTA in the United Kingdom and Ireland; UEFA Champions League football in Germany and Italy; Roland-Garros and Ligue 1 in France; Australian Swimming globally; and New Zealand Cricket in India. While availability varies by marketplace, fans can also subscribe to streaming services such as Eurosport, FOX Sports (Mexico), LaLigaTV, MLB.TV, NBA League Pass, NBA TV, and Premiere FC (Brazil) through Prime Video Channels. This is in addition to a selection of Amazon Original All or Nothing sports docuseries including All or Nothing: Tottenham Hotspur, All or Nothing: Juventus, All or Nothing: Toronto Maple Leafs, and All or Nothing seasons with the NFL’s Arizona Cardinals, Los Angeles Rams, Dallas Cowboys, Carolina Panthers, and Philadelphia Eagles, as well as the NCAA’s Michigan Wolverines football team.
About Amazon
Amazon is guided by four principles: customer obsession rather than competitor focus, passion for invention, commitment to operational excellence, and long-term thinking. Amazon strives to be Earth’s Most Customer-Centric Company, Earth’s Best Employer, and Earth’s Safest Place to Work. Customer reviews, 1-Click shopping, personalized recommendations, Prime, Fulfillment by Amazon, AWS, Kindle Direct Publishing, Kindle, Career Choice, Fire tablets, Fire TV, Amazon Echo, Alexa, Just Walk Out technology, Amazon Studios, and The Climate Pledge are some of the things pioneered by Amazon. For more information, visit amazon.com/about and follow @AmazonNews.
© 2010-2022 Amazon.com, Inc. or its affiliates Privacy Notice
Design by SPINX Digital
We use cookies to provide you with a better website experience. Continue browsing if you are ok with this, or click here to review our privacy policy and find out how to manage cookies.

source

Listen: Glenn Thompson Speech At Son's Gay Wedding – BuzzFeed News

“We’re just blessed, and we just want to say thank you to everyone here as part of the celebration,” said Rep. Glenn Thompson, three days after voting against codifying marriage equality in federal law.
BuzzFeed News Reporter
Rep. Glenn Thompson gives a speech in 2018 at a candidates’ forum in Tarentum, Pennsylvania.
As candles flickered on the long wooden dining tables under a ceiling decorated with large string lights, guests sat with glasses of sparkling wine before them to hear speeches to toast the two newlyweds.

But the Friday night wedding, southwest of Philadelphia, included a controversial speaker who just so happened to be the father of one of the two grooms.
Wearing a black tuxedo and black bowtie, Pennsylvania Republican Rep. Glenn Thompson toasted his son and new son-in-law, both of whom sat before him, wearing wedding white blazers. Standing next to his wife, the member of Congress told the guests how “blessed” he felt to celebrate the men’s marriage and to welcome a new member into his family. Thompson said any parent hopes and prays that their child stays healthy, finds their way, and ultimately finds “that one true love so that they have the opportunity to experience that: someone to grow old with.”
“We love it when they find their one true love, especially when they become a part of our families then. That’s what we’re rooting for,” Thompson said, praising his son for his selection of a husband.
But as some in the crowd knew — and some would later learn — Thompson had just three days earlier voted against a bill that would codify federal protections for marriage equality. He was one of 157 House Republicans who voted against the Respect for Marriage Act, which acts as a failsafe in case the Supreme Court reverses itself on marriage equality — something LGBTQ activists fear is a very real possibility following the court’s overturning of abortion protections.
After the vote, Thompson’s press secretary Maddison Stone told ​​the Centre Daily Times the bill was a “stunt” by Democrats — despite 47 Republicans breaking with Thompson and the rest of the party to vote in favor of it. “This bill was nothing more than an election-year messaging stunt for Democrats in Congress who have failed to address historic inflation and out of control prices at gas pumps and grocery stores,” Stone said.
The hypocrisy of Thompson’s vote prior to his son’s wedding was first noted by Gawker on Thursday — a day before the event took place. On Monday, NBC News then reported that Thompson had in fact attended the event after his “no” vote.
BuzzFeed News can report that Thompson not only attended the event, but he delivered a speech to celebrate the nuptials, which he called “a really good experience.” Media of the speech was provided to BuzzFeed News by a guest, who asked to remain anonymous.
“I think the word should definitely get out there,” the guest said of their decision for BuzzFeed News to publish it. “These politicians need to be exposed for who they really are.”
Below, BuzzFeed News is publishing an edited audio version and transcript of the remarks that includes most of Thompson’s speech, but removes sections where he names his son and son-in-law. (Like other news outlets, BuzzFeed News is not naming the two grooms as they are not public figures. When reached on his honeymoon, Thompson’s new son-in-law told BuzzFeed News that the couple weren’t sure if they want to say anything publicly right now.)
Stone, Thompson’s press secretary, didn’t immediately respond to a request for comment for this story, but previously confirmed to NBC and other outlets that Thompson had attended the wedding. “Congressman and Mrs. Thompson were thrilled to attend and celebrate their son’s marriage on Friday night as he began this new chapter in his life,” Stone said. “The Thompsons are very happy to welcome their new son-in-law into their family.”
Should the Supreme Court overturn Obergefell v. Hodges, the 2015 case that enshrined marriage equality nationwide, individual states would once again have the power to determine whether or not to refuse marriage licenses to two people of the same sex. The bill Thompson voted against would repeal the Defense of Marriage Act of 1996 and ensure that federal benefits and protections remain in place for couples like the lawmaker’s son, regardless of their state of residence.
We really appreciate you all being here, and being a part of that. You know, as a parent — parents have hopes and dreams, right, with their kids, from the time that they’re born and they’re creeping and crawling and walking and falling over and walking again, and all the things that they learn right through their teens and into becoming adults. We have hopes and dreams. First of all, obviously, we hope right from the beginning, it’s all about having a healthy child. But it’s about being healthy. It’s about them being healthy. We hope for safety. We’re hoping that they find their way, find opportunity, they find inspiration. And as they grow and as they get a little older, we also hope and pray they’re going to find that one true love so that they have the opportunity to experience that: Someone to grow old with. So we’re just really thankful that you’re here. It actually goes beyond that, as parents. We love it when they find their one true love, especially when they become a part of our families then. That’s what we’re rooting for. We’ve been fortunate with three sons, and [REDACTED]’s done a great job of adding to the family. Every kid showed up through cesarean section so it wasn’t all pleasant, right! So this has been a really good experience, especially for Penny, to have a new son enter the family! So we’re just blessed, and we just want to say thank you to everyone here as part of the celebration.

BuzzFeed News Reporter
David Mack is a senior breaking news reporter for BuzzFeed News and is based in New York.
Contact David Mack at [email protected].
Got a confidential tip? 👉 Submit it here

source

LIVE Transfer Talk: Man United eye Lazio's Sergej Milinkovic-Savic as Erik ten Hag scrambles for midfield depth – ESPN

Julien Laurens and Mark Ogden slam Manchester United for targeting Marko Arnautovic as a striker solution. (1:49)
The summer transfer window is here, and there’s plenty of gossip swirling about who’s moving where. Transfer Talk brings you all the latest buzz on rumours, comings, goings and, of course, done deals!
Manchester United are set to switch focus to Sergej Milinkovic-Savic as Erik ten Hag looks to acquire a centre midfielder before the transfer window closes, according to the Telegraph.
The Premier League side have been unable to land a priority midfield target this summer, having failed to agree personal terms with Barcelona‘s Frenkie de Jong, but the latest indicates that they will now step up their approach for the 27-year-old Lazio star as an alternative.
Sources have told ESPN that United have also made contact with Juventus about signing midfielder Adrien Rabiot. United believe that Juve are willing to let 27-year-old Rabiot leave and the France international would be keen to move to Old Trafford and play in the Premier League. Rabiot’s mother and agent Veronique has started preliminary talks.
Milinkovic-Savic registered 11 goals and 11 assists for Lazio last season, with form that has seen him previously linked to several Premier League sides. The Serbia star could now be tempted away from the Estadio Olimpico where he is now inside the final 12 months of his contract.
Corriere dello Sport has suggested that Maurizio Sarri’s side would be willing to consider offers in the region of €50 million to part ways with him. Among other clubs who have previously expressed interest in Milinkovic-Savic include Newcastle United, Arsenal, Chelsea and Paris Saint-Germain.
– ESPN+ guide: LaLiga, Bundesliga, MLS, FA Cup, more (U.S.)
– Stream ESPN FC Daily on ESPN+ (U.S. only)
– Don’t have ESPN? Get instant access

21.37 BST: After failing to secure a move for Alex Moreno, Nottingham Forest have now stepped up their attempts to sign Pervis Estupinan, reveals Foot Mercato.
Villarreal would be willing to part ways with the 24-year-old left-back for a fee in the region of €20m, and it is expected that the newly-promoted Premier League side will now look to reach full agreement with the Yellow Submarine.
The Ecuador international made 28 LaLiga appearances last season, where he registered one assist.
20.23 BST: As Barcelona work on offloading players before the window closes, they are continuing their efforts to move on Samuel Umtiti, writes Sport.
The 28-year-old centre-back has been informed that the Blaugrana would be willing to help contribute to some of his salary to assist his efforts in finding a new club, and it is reported that there has been interest from Olympiakos and in Turkey for his signature.
19.50 BST: Chelsea are readying a third offer to sign Leicester City centre-back Wesley Fofana, reports the Evening Standard.
It is understood that the 21-year-old is keen to make the switch to Stamford Bridge, and could be prepared to help force through a move, though Brendan Rodgers’ side are unsatisfied with the little time left in the transfer window they would have left to acquire a replacement. The Blues previously had an offer in the region of £70m rejected.
19.15 BST: In more LaLiga and Arsenal news, the representatives of Hector Bellerin are understood to be in talks with the Gunners over terminating his contract with the club, reveals El Chiringuito TV.
The player is keen to return to Real Betis where he enjoyed a loan spell with last season, and they are prepared to offer the 27-year-old right-back a deal between three and four years.
18.39 BST: Yeremi Pino has been attracting plenty of interest this week, and reports now suggest that Villarreal are preparing to receive an offer of up to €50m.
That’s according to AS, who write that Arsenal and Liverpool have both been keeping tabs on the 19-year-old right-winger, who contributed to 10 goals in 31 LaLiga appearances last season.
He’s back! 💪 pic.twitter.com/aRIhgSK8OV
17.46 BST: RB Leipzig have confirmed the signing of Timo Werner from Chelsea on a permanent deal. After failing to meet expectations during his stint at Stamford Bridge, the Germany international will return to the club he left to join the Blues in 2020.
Werner scored 23 goals in 89 appearances for Chelsea and only started 15 league games last season. Meanwhile, Thomas Tuchel will look to Raheem Sterling, Armando Broja and Kai Havertz to lead the Chelsea attack this season.
17.28 BST: Talks over a move for Fabian Ruiz are entering the final stages, understands Foot Mercato.
Paris Saint-Germain have stepped up their interest in the 26-year-old midfielder who has just one year left on his current deal, and the latest indicates that they are close to reaching full agreement over a transfer fee with Napoli.
The Ligue 1 club’s opening offer is believed to be €25m, while Gli Azzurri have been negotiating to receive closer to €28m.
17.09 BST: Nikola Vlasic is close to joining Torino on a loan with an option to make the move permanent, reports Gianluca Di Marzio.
The Serie A club can activate the clause for a fee of €15m if he impresses this season, and having pursued the 24-year-old West Ham attacking midfielder since the opening of the summer transfer window, it looks as though they are set to finally secure his signature.
16.05 BST: Liverpool are unlikely to go into the transfer market to sign cover for Thiago Alcantara despite the midfielder being ruled out for up to six weeks through injury on Tuesday.
The Spain international limped off the pitch early in the second half with a suspected hamstring strain during Saturday’s 2-2 Premier League draw at Fulham.
Thiago’s injury comes at a bad time for Liverpool, who are still without midfielders Alex Oxlade-Chamberlain and Curtis Jones.
“This isn’t a good situation. I don’t like it at all. We have to see how we react on that, but for sure not panic,” Liverpool coach Jurgen Klopp said.
The transfer window closes on Sept. 1 but Klopp indicated Liverpool will not make a rash decision regarding signing new players.
“A transfer must make sense now and in the long-term,” he said. “We have eight midfielders. We still have enough midfielders. It’s not that we lack midfielders, it’s just some of them are injured.”
– Olley: Liverpool could be left playing catch up
15.17 BST: Monaco are interested in signing out-of-favour Manchester United defender Eric Bailly, L’Equipe reports.
Bailly, 28, has endured mixed fortunes since arriving at United from Villarreal in 2016. The Ivory Coast international won the Europa League and League Cup in his first season at Old Trafford but has dropped down the pecking order under subsequent managers.
With Chelsea centre-back Malang Sarr reportedly set to join Monaco on loan, the principality club could bolster their defence further with the signing of Bailly. L’Equipe says they are seeking a loan deal for the 28-year-old with an option to buy.
14.35 BST: Graham Hunter takes a look at how European champions Real Madrid are cruising into the new season while their main rivals across the continent are playing catch-up:
As Real Madrid begin their quest to win six trophies in a single season, it’s remarkable how serene and untroubled their panorama is compared to the turbulence facing almost every domestic and European rival. Their daunting challenge begins with the UEFA Supercup here in Helsinki, against Europa League winners Eintracht Frankfurt, takes flight with the first LaLiga game (at newly-promoted Almeria this weekend) and marches onward through the Champions League, Copa del Rey, Spanish Supercopa and the FIFA World Club Championship, the journey finally ending next June.
Oh, and there’s the “small” matter of a World Cup for most of Los Blancos‘ players neatly tucked in during November and December.
But the minimum one can ask before a ball is kicked in anger is for the club, squad, first team and staff to be in a state of harmony, hunger and happiness. So it has been. There have been departures (Marcelo, Gareth Bale, Luka Jovic, Isco), but not a single important player has left the squad that conquered Spain and Europe last season. Two important, intelligent and potentially impactful signings have been added in Aurelien Tchouameni (a first-choice central midfielder for world champions France) and, as a free agent, Antonio Rudiger, a ferociously competitive defender who is in the habit of winning things for club and country.
This bucolic vista for their club will appear still more attractive to Madrid and Madridistas alike as soon as they take a lateral view at their competitors at home and around Europe.
– Stream LIVE: Almeria-Real Madrid 8/14, 4 p.m. ET, ESPN+ (U.S.)
13.51 BST: Manchester United have withdrawn their interest in Marko Arnautovic, sources have told ESPN’s Rob Dawson.
The club were in talks with Bologna about a deal for the former Stoke and West Ham striker but sources have told ESPN the Italian side’s demands scuppered any hope of an agreement.
United were also made aware of increasing fan backlash against the move, stemming from accusations Arnautovic used racist language during Austria’s game against North Macedonia at Euro 2020. Arnautovic, who is of Serbian descent, later apologised and insisted he is not racist although he was banned by UEFA for one game for his on-field outburst.
Sources have told ESPN that United initially targeted Arnautovic because of a belief the 33-year-old could be signed for a relatively low fee but Bologna’s valuation has increased since initial conversations took place. Ten Hag is keen to add a physical back-up striker to his squad and has worked with Arnautovic during his time as a coach at FC Twente in the Netherlands.
13.35 BST: Everton have announced the signing of midfielder Amadou Onana from Lille.
The 20-year-old has signed a five-year contract at Goodison Park after joining for an undisclosed fee. West Ham United had previously been linked with the Belgium international.
Onana becomes the Toffees’ fifth summer signing after James Tarkowski, Ruben Vinagre, Dwight McNeil and Conor Coady.
Amadou will wear #⃣8⃣ in honour of his hero, the late @kobebryant 💙🏀 pic.twitter.com/zSsx9BenHo
13.14 BST: Real Madrid and Atletico Madrid are battling it out for the signing of Bayer Leverkusen starlet Iker Bravo, sources have told ESPN’s Adriana Garcia and Rodrigo Faez.
Atletico were the first to show an interest in the Spain U19 and U17 forward. Madrid monitored the 17-year-old’s progress last season and have decided to join the race for his signature.
Sources have told ESPN that both clubs see Bravo as a prospect for the future and will not rush him, thus allowing him to go step-by-step and play for the reserves at first before joining up with the first team.
Madrid are confident Bravo will choose them ahead of Atletico, although the player’s entourage appreciate the effort made by Atletico after the club increased their offer for the striker. Whatever the outcome, all parties expect the deal to be finalised in the coming days.
12.35 BST: A move to Chelsea would be Frenkie de Jong‘s preferred choice if he were to leave Barcelona this summer, Mundo Deportivo reports.
Barca reached a preliminary agreement with Manchester United to sell De Jong for a total of €85m and needed the money in order to register their new signings without breaching LaLiga’s wage limits, but De Jong has so far been adamant that he wants to stay at Camp Nou.
Barca want De Jong, who is one of the highest paid players at the club, to accept a wage cut if he remains, and sources have told ESPN’s Sam Marsden and Moises Llorens that the club sent him a letter in July warning the player that the renewal contract he signed in 2020 under the previous board had alleged irregularities.
When contacted by ESPN, sources close to the player said the contract is legal and claim the letter that Barca sent was intended to put pressure on the Netherlands midfielder to either accept a move away from the Camp Nou or to accept a wage reduction.
Meanwhile, COPE radio reports that Chelsea left-back Marcos Alonso will not accept a transfer to Barcelona this summer until the Catalan club registers their new players.
Alonso’s contract at Chelsea expires in June 2023 and the Spain international is keen to play in LaLiga. The 31-year-old was close to joining Barca on loan in January 2021 but the deal did not materialise.
COPE also reports that Gerard Pique has agreed to a drastic wage cut with Barcelona to help the club register new players.
Pique, one of the highest-paid players at the club, is under contract with Barca until June 2024. The 35-year-old deferred his salary several times in the past three years to help Barca cope with the adverse financial impact of COVID.
12.11 BST: Tottenham Hotspur are closing in on a deal worth around £21m for Udinese defender Destiny Udogie, sources have told ESPN’s James Olley.
Spurs will loan the left-back back to the Serie A club for this season before he joins up with his new team in the summer of 2023.
Sources have told ESPN that Tottenham will pay an initial £15m with a series of performance-related add-ons worth in the region of £6m. A medical is expected to take place in Italy this week.
Should Udogie’s transfer be completed, he will become Tottenham’s seventh summer signing following Ivan Perisic, Richarlison, Fraser Forster, Djed Spence, Yves Bissouma and Clement Lenglet.
11.55 BST: Manchester United have registered their interest in PSV Eindhoven winger Cody Gakpo, sources have told ESPN’s Rob Dawson.
Gakpo’s representatives have been alerted to United’s interest by recruitment staff at Old Trafford but the club have yet to submit a bid.
Gakpo is valued at around £35 million by PSV, who are managed by former United striker Ruud van Nistelrooy, and sources have told ESPN they are unlikely to accept anything less for one of their star players.
The 23-year-old came through the club’s academy and scored 21 goals in 47 games last season. He has been capped seven times by the Netherlands, scoring twice.
Sources added there is interest from other Premier League clubs in Gakpo, who scored twice in PSV’s 4-1 opening-day win against FC Emmen on Saturday, but that he would be keen to move to United if terms are agreed.
11.41 BST: Eintracht Frankfurt have said that Filip Kostic will not play in the UEFA Super Cup against Real Madrid on Wednesday as he is “in final talks with a new club.”
The Bundesliga club announced the news after Sky Sports reported that Kostic was set to join Juventus in a deal worth up to €18m. Those reports also stated that Juve did not want Kostic to play in the Super Cup against Madrid.
Kostic can play as a winger or a wing-back and has made 172 appearances for Eintracht since initially joining on loan from Hamburg in 2018. The Serbia international played a key role in Eintracht’s Europa League win last season and was named the tournament’s player of the season.
11.17 BST: RB Leipzig have announced that Slovenia international forward Benjamin Sesko will join them from FC Salzburg next summer.
Sources told ESPN’s Rob Dawson earlier this week that Manchester United had been in talks to sign the 19-year-old, with Chelsea also interested, but instead he will move to the Bundesliga for the 2023-24 season.
Leipzig added in a club statement that Sesko will sign a five-year contract at Leipzig when he makes the move at the end of the season.
– Karlsen: Who could be the next big breakout star in Europe?
10.46 BST: Manchester United have contacted Real Betis about a potential deal for midfielder Guido Rodriguez, according to AS.
Rodriguez, 28, joined Betis from Club America two years ago and has made 100 appearances since then, helping them to lift the Copa del Rey last season.
The Argentina international is reportedly seen as an alternative to Juventus’ Adrien Rabiot if United miss out on his signing — sources told ESPN on Monday that United had made contact with the Serie A club about the France midfielder.
10.17 BST: Leeds United have signed goalkeeper Joel Robles on a free transfer from LaLiga club Real Betis.
The 32-year-old has Premier League experience having played for Wigan Athletic — where he won the FA Cup in 2013 — and Everton.
Robles has signed a one-year contract at Elland Road.
✍️ #LUFC is pleased to announce the signing of goalkeeper Joel Robles
09.58 BST: FC Copenhagen have signed Australia goalkeeper Mat Ryan from Real Sociedad on a two-year contract, the Danish Superliga champions said on Tuesday.
The 30-year-old, who played for Club Brugge, Valencia, Genk, Brighton & Hove Albion and Arsenal before his move to Sociedad, is keen to get as much game time as he can ahead of the World Cup in Qatar, which starts in November.
Australia captain Ryan, who has 73 caps, said in a statement he was looking forward to playing at Copenhangen, a club which has a “fantastic reputation throughout Europe.”
“The team has great quality and the ambitions for the club are high in terms of winning titles and competing in the biggest tournaments in Europe,” he said. “I will work hard to contribute to that.”
Ryan made only nine appearances in all competitions last season and is set to replace Copenhagen’s first-choice Kamil Grabara, who is sidelined with injury. Copenhagen won their 14th Superliga title last season.
09.28 BST: Aston Villa manager Steven Gerrard says he is waiting until Tyrone Mings “looks me in the eye and shows that he’s ready to play” before selecting the defender again.
England defender Mings captained for Villa last season, but Gerrard has given it to midfielder John McGinn this term. Asked if Mings was an unused substitute in the 2-0 loss to Bournemouth on Saturday because of his reaction to losing the armband, Gerrard told the BBC: “It’s a little bit of that.
“He’s also missed some training time with an injury and there’s a niggle in the background as well.
“When Tyrone is back at his best and looks me in the eye and shows that he’s ready to play, he’ll get opportunities.”
08.44 BST: Barcelona midfielder Nico Gonzalez wants to join Valencia in search of playing time, according to Mundo Deportivo.
Nico, 20, gained promotion to Barca’s first team last summer and made 27 league appearances for Barca last season, but the arrival of Ivorian midfielder Franck Kessie in this transfer window means Nico’s playing time will likely be reduced.
Barca coach Xavi Hernandez recently said Nico could help the team this season but the player has reportedly asked Xavi to be allowed to leave on loan to get regular playing time.
Valencia are eager to strengthen their midfield line and have made a request to Barca to take Nico on a season-long loan.
08.00 BST: Timo Werner is on the brink of leaving Chelsea to rejoin RB Leipzig, sources have told ESPN James Olley.
The 26-year-old is expected to arrive in Germany for a medical ahead of a permanent move thought to be worth in the region of €30m, although various performance-related add-ons are yet to be finalised.
Werner is keen to kick-start his career ahead of the World Cup in Qatar after a disappointing two-year spell at Stamford Bridge.
The Germany international joined Chelsea in 2020 for €50m, opting not to take part in Leipzig’s rescheduled Champions League quarterfinal against Atletico Madrid — delayed due to COVID-19 — to give himself more time to adapt to English football.
Werner started well, scoring eight goals in his first 12 games up to mid-November, but managed just four more for the rest of the campaign as a loss of form and confidence set in.
– Mixed performances for debutants Sterling, Koulibaly vs. Everton
– Luis Muriel is the latest player to be added to Newcastle United’s shortlist, according to Tuttosport. The 31-year-old versatile forward contributed to 17 goals in 27 appearances for Atalanta last season, and he is also understood to be on the radar of Juventus. Atalanta are reported to be considering potential alternatives if they receive an acceptable offer from the Magpies.
– Talks over a new contract have begun between Chelsea and goalkeeper Edouard Mendy, reveals the Daily Mail. While he still has three years remaining on his current deal, it is understood that the 30-year-old goalkeeper has one of the lowest salaries at Stamford Bridge, despite being deemed one of the squad’s most important players. It is expected that he will receive an increase after his impressive performances since joining from Rennes in 2020.
– Boca Juniors are set to sign former Manchester United goalkeeper Sergio Romero, reveals TYC Sports journalist Leandro Aguilera. The 35-year-old will sign a two-year deal with the club, and that will become official after he completes his medical. Romero spent last season in Serie A with Venezia, where he made 16 league appearances for the since relegated side.
– Juventus are considering moving on Moise Kean before the transfer window closes, according to Calciomercato. He made 32 Serie A appearances last season, but he is understood to have upset Massimiliano Allegri after he missed a team meeting. Atletico Madrid and PSG have both been linked with the 22-year-old forward, and the Bianconeri are hopeful that Ligue 1 side may be willing to make a move for him after his previous successful loan spell at the Parc des Princes. Kean enters the second season of a two-year loan deal with Juventus from Everton.
– Lucas Torreira has joined Galatasaray on a four-year contract from Arsenal, the clubs said on Monday. Torreira joined Arsenal in July 2018 from Sampdoria and made nearly 90 appearances for the Premier League club. He spent the last two seasons on loan — firstly at Atletico Madrid, helping them to their 2020-21 LaLiga title, and then with a return to Serie A by joining Fiorentina. The Turkish side have also completed the signing of Dries Mertens from Napoli, on a deal through 2023.

source

Songwriter Lamont Dozier, an architect of the Motown sound, dies at 81 – The Washington Post

Lamont Dozier, a Motown songwriter who was part of a hit-making trio that helped define the Detroit record label’s signature sound with an astonishing run of singles that included chart-topping hits for the Four Tops and the Supremes, died Aug. 8 at his home near Scottsdale, Ariz. He was 81.
His death was announced by Los Angeles-based publicist Jo-Ann Geffen, who did not share additional details.
Mr. Dozier and his songwriting partners, brothers Brian and Eddie Holland, formed a powerhouse collaboration that churned out hit after hit — “Heat Wave,” “Baby Love” and many more — and helped propel the careers of Motown legends such as Diana Ross, Marvin Gaye and the Isley Brothers.
The team, known as H-D-H, so thoroughly put their stamp on Motown that dozens of major artists from the era have at least one H-D-H song in their hit catalogue, including 10 of 12 No. 1 recordings by the Supremes, such as “Stop! In the Name of Love” (1965) and “You Keep Me Hangin’ On” (1966). In a six-decade career, Mr. Dozier had credits on more than 100 Top 40 hits.
Mr. Dozier said much of his inspiration for melodies and lyrics came from his boyhood in Detroit, listening to adults gab about the ups and down of love and relationships. He was often told to leave the room when the conversation moved to “grown-up talk” about sex. “But I’d still listen,” he recalled in a 2015 interview.
“All of this stuff stayed in my head for many years,” he added.
Mr. Dozier and the Holland brothers left the Motown label of impresario Berry Gordy in 1968 to start their own home for artists, Invictus Records and Hot Wax Records. Later, Mr. Dozier carved out a solo career that included writing back-to-back hits in 1970, “Give Me Just a Little More Time” performed by Chairmen of the Board, and Freda Payne’s “Band of Gold.”
Mary Wilson, founding member of the Supremes, dies
In 1988, Mr. Dozier worked with British rocker Phil Collins on “Two Hearts” for the film “Buster,” with the song earning a Grammy and an Academy Award nomination. The H-D-H team was inducted into the Songwriters Hall of Fame in 1988 and the Rock & Roll Hall of Fame in 1990.
Mr. Dozier never learned to fully read music or write scores. “I was too busy,” he once said. But he believed he developed a sense of chord structure and power from listening to his aunt, a classical pianist, practice in Detroit when he was young. He called the Motown sound, at its best, a mix of the chord progressions of classical music and the soulful energy of gospel.
“Torchy but not torchy, fun but not overproduced,” he said in a 2018 interview. “We wanted to get the same feeling of a ballad, without it being a ballad.”
Lamont Herbert Dozier was born June 16, 1941, and raised in Detroit’s Black Bottom district, which had swelled in the 1920s during the migration of Southern Blacks to northern cities in search of factory jobs and other work.
“Whatever you called it, it was the ghetto,” Mr. Dozier recalled this year after the 1966 H-D-H song “Reach Out I’ll Be There,” made famous by the Four Tops, became part of the National Recording Registry at the Library of Congress.
Watch Eddie and Brian Holland recount the songwriting years with Lamont Dozier
He began writing songs while still a teen, turning some lyrics into love letters he sold to friends, Cyrano de Bergerac-style, for 50 cents, according to the music website Songfacts. Mr. Dozier got a foot in the Motown door — first sweeping floors, and then in 1960 as a singer under the name Lamont Anthony. On 1961’s bluesy “Benny the Skinny Man,” he sang for the label Anna Records, owned by Gordy’s sister.
Two years later, he partnered with the Holland brothers, who were already making their mark in Gordy’s Motown empire. Eddie had a Top 30 hit as singer in 1961 with “Jamie,” but turned to writing lyrics because of crippling stage fright; Brian was co-writer on the Marvelettes’ No. 1 “Please Mr. Postman.” Mr. Dozier contributed both music and lyrics to the team.
H-D-H scored their first notable successes with “Heat Wave” in 1963 by Martha and the Vandellas and, a year later, “Where Did Our Love Go,” performed by the Supremes. A remarkable run was underway. The team’s body of work reads like a showcase of Motown’s greatest hits: 1964 with “Baby I Need Your Loving” and “How Sweet It Is (To Be Loved by You),” 1965 with “It’s The Same old Song,” 1966 with “I Hear a Symphony,” and 1967 with “Jimmy Mack,” “Bernadette,” “Standing in the Shadows of Love” and more.
Mr. Dozier said he found inspiration from his own life.
For 1965’s “I Can’t Help Myself (Sugar Pie, Honey Bunch),” he recalled his grandfather flirting, “a twinkle in his eye,” with local ladies in their Detroit neighborhood. “Bernadette” was the name of his preteen crush and “muse.” “Heat Wave” came from his memory of a sticky Detroit summer. “But it was always about love though — hot, cold or whatever,” he said.
Mr. Dozier liked to watch “The Honeymooners,” and “How Sweet It Is (To Be Loved by You)” is ripped right from a catch line of star Jackie Gleason: “How sweet it is.” When H-D-H was crafting the song for Gaye, they deliberately put it in a key slightly higher than Gaye’s comfort zone.
“If it was too easy, he’d get lazy,” Mr. Dozier said.
H-D-H continued to contribute to Motown into the early 1970s, but under the collective pseudonym Edythe Wayne or sometimes Edith Wayne because of legal disputes with Gordy over royalties. For their own labels, the trio’s most successful work was “Why Can’t We Be Lovers” in 1972. Mr. Dozier left the team in 1973 to pursue solo projects. He eventually settled in the Encino section of Los Angeles with his family.
His 1977 album “Peddlin’ Music on the Side” contained “Going Back to My Roots,” which was later rerecorded by the band Odyssey. H-D-H reunited for one last collaboration, writing the score for a 2009 stage production of “The First Wives Club,” a musical adapted from the 1996 movie comedy. The theater reviews were mixed, and the producers swapped the H-D-H score for another.
Mr. Dozier moved to Arizona after the death last year of Barbara Ullman Dozier, his wife of 41 years. Survivors include six children and three grandchildren.
During Gordy’s reign at Motown, the record label was run like an auto plant because that was the only other work environment the boss knew, Mr. Dozier said. The songwriters, session musicians and others had to punch a clock. As part of “quality control” each Friday, H-D-H and other songwriters had to write down their songs from the week, and Gordy and other executives would vote on the ones they liked.
The H-D-H songs were usually the winners, Mr. Dozier said.
“It was a fun time, like kids playing in a playground,” he said this year. “Everything we touched turned to gold.”

source

‘Scoob!: Holiday Haunt’ Producer Reveals Score Being Recorded After Film’s Axing: ‘Already Paid for the Stage and Musicians’ – Variety

Variety Print Plus Subscriber
Variety Print Plus Subscriber
Variety Print Plus Subscriber
Variety Print Plus Subscriber
Variety Print Plus Subscriber

By Thania Garcia
Although Warner Bros. Discovery has officially pulled the plug on release plans for “Scoob!: Holiday Haunt,” a followup to the 2020 animated film “Scoob!,” the film isn’t quite done with work yet. Producer Tony Cervone, who directed the first “Scoob!,” has revealed that the production is still recording the film’s score, despite the official cancellation.
On Saturday, Cervone shared a photo of his team working in a recording studio on Instagram.
“So what do you do when the movie is canceled, but you’ve already paid for the stage and the musicians?” Cervone wrote in the caption. “You record the damn score!”

A post shared by Tony Cervone (@tonycervone)

Although recording a score for a film that will not be released isn’t exactly an ordinary practice, axing a film after the bulk of its production has already been completed — which was the case around “Scoob!: Holiday Haunt,” according to reports — isn’t ordinary either.
Last Tuesday, Hollywood was rocked after Warner Bros. Discovery confirmed that “Scoob!: Holiday Haunt” along with “Batgirl” (a feature film adaptation of the DC Comics character) would no longer be released, either theatrically or on HBO Max. Both productions were conceived as original programming for the streamer and were eyeing release dates later this year.

“Holiday Haunt” was being directed by Bill Haller and Michael Kurinsky, while Dara Taylor (“Barb and Star Go to Vista Del Mar,” “The Tender Bar”) was credited as the film’s composer. Producer and co-writer Cervone confirmed the cancellation on Instagram.
“Yes I am afraid this is true,” he wrote. “The movie is practically finished and turned out beautifully. I am beyond heartbroken.”
Warner Bros. Discovery CEO David Zaslav was quoted addressing the controversial decision to throw out “Batgirl” during a company earnings call later last week, saying, “we’re not going to put a movie out unless we believe in it.”
“Batgirl” was set to star Leslie Grace as Barbara Gordon (aka Batgirl), with “Bad Boys for Life” and “Ms. Marvel” directors Adil El Arbi and Bilall Fallah at the helm. The production was approved back in 2021 as part of a company-wide effort at Warner Bros. to generate feature films for HBO Max. The film had already completed principal photography and was beginning test screenings.
The move to chop “Batgirl” and “Scoob!: Holiday Haunt” comes as Warner Bros. Discovery has re-adjusted the company to focus on theatrical features.
“Scoob!: Holiday Haunt” was slated for release on HBO Max on Dec. 22. The film reportedly cost around $40 million to make. The company will reportedly look to take a tax write-off on the incomplete production.
“Holiday Haunt” would’ve starred the voice of Frank Welker as Scooby-Doo, alongside Iain Armitage, Ariana Greenblatt, Pierce Gagnon, Mckenna Grace, Michael McKean, Mark Hamill, Andre Braugher and Jim Meskimen.
Correction: A previous version of this article identified Tony Cervone as a director of “Scoob! Holiday Haunt.” He is a producer. The piece has been amended.
A Variety and iHeartRadio Podcast
The Business of Entertainment

source

Various Jewelry Stolen From Master Bedroom: Northbrook Police Blotter – Patch

NORTHBROOK, IL — The following information comes from the Northbrook Police Department and court records as a record of incidents reported to police and those arrested on criminal charges, which represent accusations by the state that are often dropped or reduced. Updated information may be available from the Cook County Clerk of the Circuit Court. Everyone arrested is presumed to be innocent unless found guilty in court beyond a reasonable doubt.
DRIVING WHILE LICENSE SUSPENDED/REVOKED

RETAIL THEFT

Spiro G. Houmpavlis, of Skokie, is accused of retail theft, and was arrested at 2:44 p.m. Aug. 2, in the 1400 block of Lake Cook Road. A court date of Sept. 1 was assigned.
POSSESSION OF CONTROLLED SUBSTANCE
Anjelika Accardi, of Northbrook, had an arrest warrant for failure to appear in court for possession of a controlled substance, and was arrested at 10:30 p.m. Aug. 2, in the 2000 block of Beechnut Road. A court date was pending at the time of the report.
LAST WEEK: Theft, Cases Of Battery On Skokie
RESIDENTIAL BURGLARY
IDENTITY THEFT
A person called police at 11:39 a.m. July 31, to report three people attempted to use her information to unlawfully make purchases utilizing her credit account at a business in the 1500 block of Lake Cook Road. The loss prevention staff contacted the complainant to verify if these individuals could use her name and credit account. After the complainant stated she didn’t know them, the store staff asked them for their identification. The subjects left the store in a dark-colored sport utility vehicle. The subjects are described as:
VANDALISM
An employee contacted police at 10:16 a.m. Aug. 1, to report various graffiti in several places in a park in the 1400 block of Maple Avenue.
RETAIL THEFT
A person reported on July 29 three people (two males and a female) entered a store and stole merchandise in the 1000 block of Willow Road. The offenders were seen driving away in a black Acura sedan.
THEFT FROM AUTO
A person called police to report that between July 31 and Aug. 1, someone stole the catalytic converter from her 2006 Honda CR-V while it was parked in the parking lot in the 1000 block of Skokie Boulevard.
THEFT
A complainant called police at 1:06 p.m. Aug. 3, to report the company she worked for was the victim of a wire fraud. She said the company received an order in June. The customer later claimed he over paid via wire transfer and asked that they refund him the overpayment. After the company refunded the money, it was found that the original payment was fraudulent.


Get more local news delivered straight to your inbox. Sign up for free Patch newsletters and alerts.
To request removal of your name from an arrest report, submit these required items to [email protected].

source

Learn how to setup Cubase for podcast recording

'Better Call Saul' Season 6 Final Episodes Directors Went to 'Master Class' in Black & White – Showbiz Cheat Sheet


Binge-worthy entertainment news and celebrity interviews
Better Call Saul Season 6 had the show’s first all black and white episode with “Nippy.” The subsequent episode, “Breaking Bad,” had significant black and white segments too. As established by previous season premieres, Jimmy McGill (Bob Odenkirk)’s future life as Gene Takovic is in black and white. Now that it’s become a focal point of the Better Call Saul final episodes, the directors went to black and white school
Director Michelle MacLaren and cinematographer Paul Donachie appeared on the Better Call Saul Insider podcast on July 26 after “Nippy” aired. They described the black and white process of Better Call Saul’s final episodes. Better Call Saul Season 6 airs Mondays at 9 p.m. on AMC.
MacLaren was a regular director on Breaking Bad too. She credited co-creator Peter Gould with schooling the directors of the Better Call Saul final episodes how he wanted the black and white to look.
RELATED: ‘Better Call Saul’ ‘Breaking Bad’ Episode Changed Saul Goodman’s Introduction to Walter White
“Peter, you did something really cool at the very beginning of this when you knew that you were going into black and white,” MacLaren said on Better Call Saul Insider. “You got all the directors who were shooting the last few episodes together. And you gave us a master class on shooting black and white film, and it was phenomenal.”
MacLaren didn’t specify the movies Gould showed. But they were enough to convey what he wanted from the Better Call Saul final episodes. MacLaren explained how they used the black and white to convey Gene’s journey. 
Stay on your toes, the penultimate episode of #BetterCallSaul premieres tomorrow. pic.twitter.com/IOLgx7PR8f
RELATED: ‘Better Call Saul’ ‘Breaking Bad’ Episode Draws Subtle Parallels Between Walter White and Saul Goodman
“We took a deep dive that Peter had put together from all these clips of different movies that were decades old,” MacLaren said. “And talked about what you liked about them and what you were hoping for this. The cinematographers were there too. And Paul Donachie our DP really did a phenomenal job with playing with the light and the dark, and having Jimmy/Gene walk from the darkness into the light. And vice versa, depending where we were in the story.”
Donachie discovered filming Better Call Saul in black and white required different skills. To say filming is a bit of a misnomer too. Donachie worked with the digital cameras to create black and white for the final episodes of Better Call Saul.
I found some technical differences. What I found was black and white, funnily enough, you need more light. I have no idea [why] to be honest. I’m not a technical person at all but I just needed more light in there to see stuff. The rating of the camera’s the same. You’re actually recording it raw which is not in color, it’s not in black and white, you’re just recording ones and zeros. It’s fundamentally having a color picture which is turned into a black and white one. Other than that really it was just interesting to try to create more shadow and actually trying to create darker and lighter areas, things that were a bit more contrasty, having deep blacks.
RELATED:

source

Fractional neural sampling as a theory of spatiotemporal probabilistic computations in neural circuits – Nature.com

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.
Advertisement
Nature Communications volume 13, Article number: 4572 (2022)
460 Accesses
3 Altmetric
Metrics details
A range of perceptual and cognitive processes have been characterized from the perspective of probabilistic representations and inference. To understand the neural circuit mechanism underlying these probabilistic computations, we develop a theory based on complex spatiotemporal dynamics of neural population activity. We first implement and explore this theory in a biophysically realistic, spiking neural circuit. Population activity patterns emerging from the circuit capture realistic variability or fluctuations of neural dynamics both in time and in space. These activity patterns implement a type of probabilistic computations that we name fractional neural sampling (FNS). We further develop a mathematical model to reveal the algorithmic nature of FNS and its computational advantages for representing multimodal distributions, a major challenge faced by existing theories. We demonstrate that FNS provides a unified account of a diversity of experimental observations of neural spatiotemporal dynamics and perceptual processes such as visual perception inference, and that FNS makes experimentally testable predictions.
Humans and other animals operate in a world that is noisy and ambiguous. Moreover, it has been widely observed that the firing activity of neural systems is inherently stochastic and neural responses to stimuli exhibit large variability within and across trials1,2,3. Such exogenous and endogenous randomness naturally leads to the view that neural computations are carried out in a probabilistic way4,5,6,7. Indeed, a host of brain functions from sensory processing8,9 and cognitive tasks10,11 to motor behaviors12 have been successfully characterized from the perspective of probabilistic inference. The success of probabilistic accounts of these brain functions raises the fundamental question of how neural activity represents probability distributions and how neural circuits implement probabilistic inference based on that representation.
Two prominent types of models have been proposed to understand the neural basis of probabilistic representations and computations. One type is based on neural population activity that encodes the parameters of probability distributions (e.g., probabilistic population codes) such as the variance of Gaussian distributions4,13. The other type of model employs sampling-based probabilistic representations14,15,16,17,18,19. Probabilistic representations in these models are mainly based on the variability or fluctuations of neural activity with Poisson or Gaussian statistics. However, empirical evidence has increasingly demonstrated that neural fluctuations occur at multiple scales with heavy-tailed, non-Gaussian statistics20,21, and that such fluctuations unfold in both time and space to give rise to rich, complex spatiotemporal dynamics of neural circuits22,23.
Here we argue that these rich spatiotemporal dynamics enable a new type of sampling-based probabilistic neural computation. This probabilistic computation offers a unique perspective on the functional role of spatiotemporal neural dynamics and provides a solution to a long-standing challenge of reliably sampling complex probability distributions such as multimodal distributions. Sampling multimodal distributions is fundamentally important for processing natural environments replete with multiple, distributed salient patches (i.e., modes)24,25, and for learning and inference26. This form of processing remains a major challenge in previous studies, because Gaussian fluctuations or noise used in existing models essentially implement Brownian-motion-based Markov chain Monte Carlo (MCMC) sampling and its variants14,15,17. Due to the lack of ‘large jumps’ in Brownian motion (Fig. 1a), when faced with multimodal distributions with far-apart modes, neural samplers are unable to traverse through low-probability regions; such samplers are thus prone to be trapped in one local mode and lack the capacity of freely switching from one mode to another27.
a Typical sample path of a Brownian motion and that of a Lévy motion. The latter consists of clusters of short step sizes that are intermittently interspersed by long jumps. In contrast, the Brownian motion lacks such long jumps. Color indicates time. b Schematic diagram of the two-dimensional E-I spiking neural circuit. The circuit consists of recurrently connected excitatory and inhibitory neurons spanning a two-dimensional feature space (top layer), receiving multiple feedforward inputs (bottom layer). The spiking activity pattern (dots) emerging from the circuit exhibits rich, complex spatiotemporal dynamics. As indicated by the red arrow, the localized activity pattern intermittently switches between different parts of the neural circuit (black circles) and can implement sampling-based probabilistic computations. c Schematic diagram of a generative process for sampling-based probabilistic inference in our model.
We illustrate our probabilistic computation theory through a biophysically realistic, spatially-extended spiking neural circuit (Fig. 1b), demonstrating that population activity patterns (i.e., neural ensembles) emerging from the circuit possess large intrinsic fluctuations at multiple spatial and temporal scales as empirically observed. Rather than Brownian motion, the activity patterns propagating across the neural circuit with such large fluctuations exhibit clusters of short step-sizes that are intermittently interspersed by long jumps (Fig. 1a), with the movement step sizes showing heavy-tailed, non-Gaussian (Lévy) statistics28. These intermittent, long jumps inherent in Lévy motion enable the activity patterns to adaptively and freely switch between different modes of multimodal distributions, thus sampling these distributions with great efficiency. Besides these heavy-tailed Lévy motions in space, our sampling approach exploits certain temporal oscillatory components that play a fundamental role in speeding up the sampling process. We demonstrate that our sampling-based representation accounts for key response properties of neural circuits such as the reduction of neural variability after stimulus onset3,29 and theta oscillations (3–8 Hz) accompanied by 1/f activity as widely observed during environmental sampling tasks30, and that the sampling dynamics of the neural circuit are the underlying mechanism for perceptual switching31.
We further elucidate that the key spatial and temporal properties of the sampling-based representation implemented in the circuit model can be characterized by a stochastic differential equation with a fractional order derivative, which generalizes the notion of differentiation to fractional orders and captures heavy-tailed Lévy motions occurring at multiple scales28. We thus develop a mathematical model of our sampling approach; due to the fractional nature revealed by this mathematical model, we term our sampling approach fractional neural sampling (FNS). Based on this mathematical model, we illustrate the essential algorithmic properties of FNS for efficiently sampling multimodal distributions and further validate predictions from the mathematical model in our spiking neural circuit, thus revealing how and why FNS works at both the circuit implementation and algorithmic levels.
FNS-based probabilistic inference produces estimates about the mean and variance of probability distributions that match those of optimal Bayesian inference. However, due to the presence of small movements occasionally interrupted by large ones in FNS, it naturally generates switching-like behaviors and the resultant multi-modes of the estimate probability distributions. Our FNS-based inference thus provides a mechanistic account of why there exists bimodality in estimate distributions, as observed in a recent psychophysics study of visual perception inference32; this experimental result is not explained by conventional models of probabilistic inference. Our FNS-based inference model further makes quantitative predictions about how statistics of perceptual estimate distributions are related to stimulus contrast; these predictions are consistent with reanalysis of the existing experimental data. Dynamical switching between different activity patterns representing either multiple external sensory inputs33 or multiple choices in decision making34,35 have been widely observed, suggesting that FNS-based probabilistic computations could be of general applicability to understanding brain functions ranging from sensory processing to decision making.
We first illustrate how FNS works based on a biophysically realistic circuit model of spiking neurons and then reveal the circuit mechanism underlying the emergence of key spatial and temporal properties of FNS. Second, we formulate FNS by using a mathematical model derived from fractional diffusion formalisms, based on which we elucidate the computational properties of FNS for sampling multimodal distributions. We then validate these properties in the spiking neural circuit model. Finally, we illustrate how FNS-based probability inference can be implemented in the spiking neural circuit with the prior embedded in recurrent synaptic weights, and demonstrate that our model provides a novel account of visual perceptual inference as observed in experimental studies.
We consider a biophysically realistic spiking neural circuit model of excitatory and inhibitory neurons that incorporates experimentally established properties of the cortex, such as distance-dependent synaptic connectivity36,37 and correlated excitatory and inhibitory synaptic inputs38 (Fig. 1b, see “Methods”). The neural circuit spans a two-dimensional feature space39 with its x and y coordinates representing feature values such as orientation (angle) and color (hue), respectively, both ranging from −π to π. The circuit model exhibits a rich repertoire of dynamical activity states including asynchronous activity (i.e., disordered state) and localized propagating waves (i.e., ordered state)40, depending on the relative strengths of synaptic inhibition and excitation (characterized by an I–E ratio ξ; see “Methods”). Near the phase transition (ξ = ξc = 3.4) between these two states, localized activity patterns with rich, complex spatiotemporal dynamics emerge and behave like random walkers, exhibiting clusters of small movement step-sizes that are intermittently interspersed by long jumps (Fig. 2a). Such intermittent motion of the localized patterns can be characterized by Lévy motion (Fig. 1a), a type of non-equilibrium motion that has been shown to be essential for animals to optimally search for spatially distributed food41,42, for T-cells to efficiently find target pathogens in brain explants43, and for optimally transporting energy in turbulent fluids44. Notably, it has been shown that Lévy motion underlies the propagation of gamma (30–100 Hz) burst patterns in the MT area of marmoset monkeys45, and that hippocampal sharp wave ripples exhibit random movements with occasional long jumps46.
a Snapshot of the localized activity pattern when the I–E ratio ξ = ξc = 3.4. Dots denote spikes during a 5 ms period. The small circle marks the center-of-mass (CoM) of the pattern and the large circle represents one standard deviation of the spike locations. The line shows the CoM trajectory of the pattern over the previous 80 ms, with color indicating time. b Mean squared displacement (MSD) of the trajectory of the activity pattern as a function of time lag. The red line indicates a power-law fit, MSD(Δt) ∝ Δtη, with the diffusion exponent η = 1.18. c Distribution of trajectory increments. The red line indicates a fitted symmetric α-stable distribution, with the tail index α = 1.28. Inset: distribution of the trajectory increment on a log–log scale. d A typical sample trajectory overlaid on top of the sampled probability density. Large jumps with lengths greater than 0.25 are highlighted with connected green dots. e Autocorrelation of the sample trajectories for different stimulus contrast levels. Averaged across 100 trails of 10 s duration. f Mean squared error of the sample mean decreases with time lag, with a small baseline due to quenched noise (randomness in synaptic connectivity). g Dependence of the sample variance on input contrast; shades indicate the standard deviation of the sample variance across 20 trials.
To quantify the Lévy motion of the localized activity pattern (Fig. 2a), we calculate its mean-squared displacement and the distribution of the increment of the pattern’s movement as in ref. 43. We track the pattern over time and calculate its center of mass (CoM), ({hat{{{{{{{{bf{s}}}}}}}}}}_{t}=({hat{x}}_{t},{hat{y}}_{t})) (see “Methods”). As shown in Fig. 2a, the CoM trajectories exhibit random diffusive properties with variable step sizes, resulting in small-movement clusters occasionally interspersed by long jumps to new locations. The mean-squared displacement of the pattern can be calculated based on the CoM as, ({{{{{{{rm{MSD}}}}}}}}({{Delta }}t)=langle parallel {hat{{{{{{{{bf{s}}}}}}}}}}_{t+{{Delta }}t}-{hat{{{{{{{{bf{s}}}}}}}}}}_{t}{parallel }^{2}rangle), where Δt is the time lag. As shown in Fig. 2b, the mean-squared displacement is linear on a log–log scale, indicating that it is a power function of Δt, such that MSD(Δt) ∝ Δtη. The diffusion exponent η determines the type of the random motion: a value of η = 1 indicates Brownian motion, commonly used for implementing MCMC sampling47; a value of η > 1 indicates a superdiffusive process and η < 1 indicates a subdiffusive process. We find that the localized activity pattern emerging in our network has η = 1.18 (Fig. 2b), indicating that its movement is superdiffusive. We further examine the distribution of the increment ({{Delta }}{hat{x}}_{t}={hat{x}}_{t+{{Delta }}t}-{hat{x}}_{t}) of the pattern’s CoM trajectory by fixing the time interval Δt = 15 ms and find that it exhibits a heavy tail. The increments can be fitted to a symmetric Lévy stable distribution with index 1 < α ≤ 2 (see “Methods”); using maximum likelihood, we find that α = 1.28 (Fig. 2c). The tail of this distribution asymptotically follows a power law (p({{Delta }}hat{x}) sim|{{Delta }}hat{x}{|}^{-1-alpha }), where α is often referred to as the tail index. The heavy-tailed, power-law distribution of the increment and the corresponding superdiffusion are the characteristic features of Lévy motions28,48.
We next illustrate how the localized activity pattern (i.e., assemblies of neurons) with Lévy motion is able to implement FNS. To provide a conceptual understanding of the computational goal of FNS, consider the probabilistic generative process shown in Fig. 1c. A stimulus x (such as an image) is defined in terms of its latent feature s (such as its orientation and color) and a global contrast level c. For clarity, we denote the true value of the stimulus feature in a particular trial as s* to distinguish it from the generic latent variable s. The stimulus x evokes a sensory response r (spike counts). We assume that the sensory response depends on the latent feature in the form of a probabilistic population code; that is, the sensory response ri of each neuron i follows an independent Poisson distribution with a bell-shaped firing rate profile centered at s* with height proportional to the stimulus contrast c (see Eq. (10) in “Methods”). Thus, the sensory response r conveys both information about the latent feature s* as well as the strength of the sensory evidence (proportional to stimulus contrast). The sensory response is used as the feedforward input to the recurrent circuit, where probabilistic inference is performed by combining the sensory evidence conveyed by the feedforward input and the prior embedded in the recurrent circuit. Mathematically, this is expressed with Bayes’ rule
where p(s∣r) and p(s) represent the posterior and the prior distributions, respectively. In our FNS approach, a sample (hat{{{{{{{{bf{s}}}}}}}}}) from the feature space is represented by the instantaneous CoM of the localized spiking pattern in the recurrent circuit. The distribution resulting from the random motion of the localized activity pattern then approximates the posterior distribution.
In this section, we first consider the simple case of sampling a unimodal distribution corresponding to sensory evidence with a flat prior to illustrate the key spatial and temporal properties of FNS. The neural circuit receives a unimodal feedforward input centered at the true value of the stimulus feature s*, with its strength proportional to the stimulus contrast level c, as described by Eq. (10) (“Methods”). Figure 2d shows a typical sample path traced by the CoM of the localized spiking pattern, which exhibits complex spatial and temporal dynamics. Spatially, the sampling path exhibits the signature of Lévy motion, i.e., the presence of occasional large jumps in space (Fig. 2d). Similar to the case of spontaneous activity, the distribution of the sample path increment is characterized by a heavy tail. Temporally, the sample path displays an oscillatory component as indicated by its autocorrelation (Fig. 2e), which can be fitted as a function of the form (exp (-{{Delta }}t/tau )cos (2pi f{{Delta }}t)) with the decay constant τ ≈ 36 ms and the oscillation frequency f ≈ 5.5 Hz. The frequency of the oscillation is higher for the stronger contrast level (Fig. 2e). It is interesting to note that this frequency is in the range of theta oscillations (3–8 Hz), an elementary oscillatory component widely observed during environmental sampling tasks such as spatial attention sampling30 and whisking in rodents49. We will come back to this point in “Discussion”.
We next calculate the convergence speed of statistical estimates based on these samples toward the true value of the stimulus feature; in particular the convergence of the sample mean, ({bar{{{{{{{{bf{s}}}}}}}}}}_{T}=frac{1}{T}intnolimits_{0}^{T}{hat{{{{{{{{bf{s}}}}}}}}}}_{t}dt). As shown in Fig. 2, the mean-squared error of the sample mean, ({mathbb{E}}[parallel {bar{{{{{{{{bf{s}}}}}}}}}}_{T}-{{{{{{{{bf{s}}}}}}}}}{*}{parallel }^{2}]), with s* denoting the true value of the stimulus feature, decreases to half-maximum at THM = 128 ms for contrast c = 1 and THM = 81 ms for c = 2, which are around the same order as the adaptation time constant τK = 80 ms and are several times the membrane time constant (τm = 15 ms). The rate of convergence is faster for the higher contrast level, as consistent with the observation that the frequency of oscillation increases with contrast (Fig. 2e). This result indicates that the activity pattern emerging from the circuit implements sampling-based representation of the stimulus features, and that the sampling process is quite efficient.
As the stimulus contrast level represents the strength of sensory evidence, the increase in contrast should lead to a reduction of the sample variance, as suggested by the generative model in Eq. (1) as well as by existing studies13. We find that this property is indeed satisfied in FNS: as shown in Fig. 2g, the sample variance (frac{1}{T}intnolimits_{0}^{T}{({hat{x}}_{t}-{bar{x}}_{T})}^{2}dt) (calculated over a time period T = 10 s) decreases with stimulus contrast. From the dynamics aspect, the reduction of sample variance with contrast is primarily due to the modulation of the random motion of the activity pattern by the feedforward input; that is, as the contrast level of the input increases, the CoM of the pattern becomes more concentrated around the center of the input.
We now demonstrate that the sampling dynamics of FNS capture key response properties of the cortex including the reduction of neural firing variability3,29. To this end, we calculate the spike-count Fano factor for individual neurons by using a fixed time window of Δt = 100 ms. The Fano factor following stimulus onset has a pronounced drop compared to the Fano factor during spontaneous activity3. We find that this reduction of Fano factor displays a U-shaped dependence on the stimulus feature (orientation) as found in the middle temporal area of monkeys29, with the largest reduction occurring when the stimulus orientation is equal to the neuron’s preferred orientation (Fig. 3a). As shown in Fig. 3b, the Fano factor is greater than one for the spontaneous activity (c = 0) and is reduced by an amount largely proportional to stimulus contrast, with the exception of low contrast levels (c ≤ 0.3).
a Fano factor tuning of neural responses for varying stimulus feature values s relative to the neuron’s preferred stimulus spref for contrast level c = 1 (dots). Dashed line corresponds to spontaneous activity when c = 0. Fano factor is calculated using spike counts over 100 ms time windows. b Trough of the Fano factor in (a) as a function of the contrast level c of the feedforward input. Shades indicate the standard deviation across different neurons. c Theta oscillations revealed by the power spectral density (PSD) of the population firing rate with a peak frequency at 6.2 Hz. Solid line and shade indicate the average and standard deviation across 10 trials, respectively, each with a duration of 10 s. Inset: the PSD on a log–log scale reveals a power-law tail with an exponent of −2.28 (dashed line, offset for visibility).
We also investigate how the Fano factor in the recurrent circuit depends on the width of the feedforward input. As shown in Fig. S1a, when the width of the input is smaller than that of the receptive field, the Fano factor decreases with the input width; the converse is true when the input width is larger than the receptive field. It has been shown that in the primary visual cortex of monkeys, Fano factor exhibits a similar trend as the stimulus size varies (see Fig. 2f of ref. 50). In addition, we find that as the input width increases, the uncertainty of the sampled distribution changes in the similar way as the Fano factor (Fig. S1b). This correspondence between the uncertainty of neural responses and that of the sampled distribution is due to the reduction (or increase) in the variability of the instantaneous firing rates of neurons as the variance of the samples (represented by the CoM of the spiking pattern) decreases (or increases).
Another key feature of neural responses arising from the sampling dynamics is that neural population firing rate possesses a theta oscillatory component (3–8 Hz), as indicated in its power spectral density (PSD) (Fig. 3c). Interestingly, these oscillations ride on top of a 1/f arrhythmic component. Certain oscillatory components such as theta accompanied by 1/f-like activity have been widely observed in the cortex20,21; they are particularly relevant for cognitive functions such as visual-spatial attention51,52. These results provide further neurophysiological validity of our circuit model of FNS.
We next elucidate the circuit mechanism underlying the emergence of the sampling dynamics in our circuit model. More specifically, we aim to pinpoint the origin of the two key features of FNS: the fractional Lévy motion and the theta oscillatory component.
The circuit model exhibits a rich repertoire of dynamical activity states, ranging from the asynchronous to propagating wave states40. Particularly, by varying the I–E ratio (ξ), it has shown that around the transition state between the asynchronous (i.e., disordered) and localized propagating wave (i.e., ordered or coherent) states, the circuit model explains a range of nonlinear neural response properties; this state transition has been characterized by calculating the susceptibility and the branching indices of neural spikes40. Figure 4a shows that when the I–E ratio ξ is large, the network exhibits an asynchronous state without any structured patterns (State III, Fig. 4b), in which neural correlations (see “Methods”) are low. On the other hand, when the I–E ratio is small, coherent patterns emerge from the circuit in the form of a localized propagating wave; this wave pattern propagates across the neural circuit with a relatively smooth and regular trajectory (Fig. 4c). In this state (State I), neural correlations are much greater than those in the asynchronous state. In the transition regime (State II) between these two states (State I and State III), neural correlation changes rapidly as ξ varies. Only in State II, the CoM of the pattern exhibits fractional Lévy motion with small steps occasionally interrupted by long jumps (Fig. 1a)51, as indicated by a tail index of α < 2 with the trough of α coinciding with the transition point at ξ = ξc = 3.4. This value of I–E ratio (ξc = 3.4) is quantitatively consistent with that measured in the visual cortex of awake mice53. It is interesting to note that this dynamical mechanism underlying Lévy motion in our circuit model is similar to other complex physical systems whose critical phase transitions are essential for the emergence of Lévy motion54.
a Phase diagram of the emergent activity states in the spiking neural circuit as the inhibition-to-excitation ratio ξ varies. Neural correlations (blue line) of spike counts reveal a state transition from the asynchronous state (State III, weak correlation) to propagating wave state (State I, strong correlation). Only within the transition regime (State II), the increment distribution of the trajectory of the localized spiking pattern follows the α-stable distribution with its power-law tail index α smaller than 2, indicating the presence of fractional Lévy motion. b, c Snapshots of the spiking pattern for State I and State II featuring asynchronous activity (ξ = 4.25, 15 ms time window) and a localized propagating wave (ξ = 2.55, 2 ms time window), respectively. Red curve indicates the CoM trajectory of the propagating wave. d When adaptation is removed, the histogram of the sample increments (solid line) still exhibits a power-law tail (dashed line). e Sample path autocorrelation shows the absence of anticorrelation as a result of removing neural adaptation. f Mean squared error of the sample mean decreases slowly with the time lag, with a baseline due to quenched noise (randomness in synaptic connectivity).
Previous studies have shown that negative feedback mechanisms such as spike-frequency adaptation or depression are essential for the emergence of oscillations55,56. Our realistic spiking network model also incorporates spike-frequency adaptation in the form of slow potassium currents, which could be the origin of oscillation in our model. To test this, we remove the adaptation completely from the network by setting the potassium current in Eq. (5) to zero. We use the spontaneous activity to re-calibrate the inhibitory synaptic strength by a factor of 1.2 so that the neural dynamics is restored to the transition regime between the asynchronous and localized pattern states. Within this regime, the increments of the CoM of the localized activity pattern still follow a symmetric Lévy stable distribution, with the tail index α = 1.19 (Fig. 4d), indicating that the pattern moves in a fractional Lévy manner. However, we find that in the circuit without adaptation, the sample autocorrelation function of the sample path decays exponentially as (exp (-{{Delta }}t/tau )) with τ = 13–19 ms when the pattern samples the unimodal distribution; it does not show any feature of oscillation or anticorrelation (Fig. 4e). This lack of theta oscillations in the circuit model without adaptation results in a slower decay (time-at-half-maximum ΔtHM = 317 ms for c = 1 and ΔtHM = 220 ms for c = 2) of the mean squared error of the mean estimate (Fig. 4f) than for the case with oscillations (Fig. 2f), indicating that theta oscillations play a role of speeding up sampling of FNS. These results indicate that the oscillatory component of FNS originates from neural adaptation in our model.
We next develop a mathematical model to gain further theoretical insights into the probabilistic sampling processes implemented in our spiking circuit model. Based on this mathematical model, we reveal the unique computational properties of FNS through highlighting the functional roles of temporal oscillations and Lévy motion.
As demonstrated above, the spiking pattern behaves like a random walker that exhibits far richer spatiotemporal dynamics than Brownian motion: the pattern exhibits occasional long jumps in space, a characteristic feature of Lévy motion, and an oscillatory component in its CoM. To model a random walk with these features, we use a stochastic differential equation (SDE) driven by Lévy motion with an auxiliary momentum term
where ({hat{x}}_{t}) is the CoM of the activity pattern, vt is an auxiliary variable representing momentum, β is the damping coefficient, b(x) is a drift term related to the probability landscape, γ is the strength of the noise, and ({L}_{t}^{alpha }) is the Lévy motion whose step sizes over a time period Δt follow a symmetric Lévy stable distribution ({{{{{{{mathcal{S}}}}}}}}alpha {{{{{{{mathcal{S}}}}}}}}(alpha,{{{Delta }}t}^{frac{1}{alpha }})) possessing a power-law tail with a tail index 1 < α ≤ 2 (see “Methods”)28. It has been shown that the momentum term vt is responsible for generating temporal oscillations in the trajectory of the random walker57, with the frequency of the oscillations controlled by the damping coefficient β. Thus, the mathematical model (Eq. (2)) is able to capture the essential dynamical features of the localized pattern, i.e., the Lévy motion and the temporal oscillations. For analytical tractability we consider the case with one spatial dimension.
As the trajectory of a random walker evolves over time, its probability density converges towards a stationary distribution (target distribution) π(x). Therefore, the trajectory of the random walker, after a sufficient burn-in time, can be considered as approximate samples from π(x). In the case of Eq. (2), the stationary distribution π(x) is the stationary solution (marginalized by treating momentum v as a nuisance variable) to the corresponding fractional Fokker-Planck equation, and it is related to the drift term b(x) by (see Method for mathematical derivations of this result)
where ({{{{{{{{mathcal{D}}}}}}}}}_{x}^{alpha -2}) is the partial Riesz fractional derivative. Using Eq. (3), we can determine the drift term b(x) required for producing samples from any desired target distribution π(x).
This mathematical model provides a simple yet effective theoretical framework for revealing the key computational properties of FNS. Without loss of generality, we set the tail index in Eq. (2) to be α = 1.2, similar to the tail index characterizing the Lévy motion of the spiking patterns emerging in the neural circuit model, and set β = 1 to capture the oscillatory aspect of FNS; other values of 1 < α < 2 and β > 0 would generate qualitatively similar results. We then compare the computational performance of this default model with three other sampling processes: (1) Sampling without oscillation (β = 0) but with fractional Lévy motion (α = 1.2). (2) Sampling with oscillation (β = 1) but without fractional Lévy motion (α = 2), which is the standard Hamiltonian Monte Carlo (HMC) sampling57. Such Hamiltonian dynamics have been previously mapped to a neural network with excitatory and inhibitory populations58 for performing efficient sampling. (3) Sampling with neither oscillation nor fractional Lévy motion (α = 2 and β = 0). For this case, the sampling process reduces to the standard MCMC driven by Brownian motion47, called Langevin sampling, which is used as a general-purpose algorithm in machine learning; it has been proposed that neural networks may implement such sampling14,15,16,59.
Typical sample paths of the four cases for sampling a standard normal distribution with zero mean and unit variance are shown in Fig. 5a, each exhibiting distinct features. Immediately noticeable are the jumps in the sample paths of the two sampling processes powered by the fractional Lévy motions, but are absent in their non-fractional counterparts. Despite the drastic differences in the sample path structure, all four cases are able to produce samples from the target distribution, as seen through the agreement between the sample histogram and the target distribution (right panels in Fig. 5a). The finer temporal structures of these sampling processes are further revealed through calculating the autocorrelation function of the sample path, (langle {hat{x}}_{t+{{Delta }}t}{hat{x}}_{t}rangle), as we have calculated for the spiking pattern in the circuit model. As shown in Fig. 5b, the autocorrelation functions of the two cases without the oscillatory component (β = 0) decay exponentially to zero. However, those with the oscillatory component (β = 1) drop to zero quickly and exhibit negative-going lobes. This indicates that successive samples rapidly decorrelate, speeding up the sampling processes and thus resulting in a faster convergence to the target distribution. As shown in Fig. 5c, the mean-squared error decays significantly faster for the two cases with the oscillatory component (β = 1), with a time-at-half-maximum of 1.42 for α = 1.2 and 1.70 for α = 2, than that without oscillation (β = 0), with a time-at-half-maximum of 1.95 for α = 1.2 and 2.51 for α = 2. These results are consistent to the observations in the spiking neural circuit model in Fig. 2, and show that the temporal oscillatory property of FNS plays an essential role in improving sampling speed. Note that the cases with the momentum (β > 0) and without the momentum (β = 0) in the mathematical model correspond to neural sampling in the circuit model with and without adaptation, respectively. In the mathematical model, the momentum pushes successive samples further apart, giving rise to the oscillatory behavior and accelerating the sampling speed. In the neural circuit model, adaptation plays a similar role by pushing the localized activity pattern away from its present location, thus speeding up sampling.
a Sampling processes driven by fractional Lévy motion (α = 1.2, top two panels) exhibit long jumps, in contrast to their non-fractional counterparts (α = 2, bottom two panels). Right panels: sample histograms agree with the target distributions (dashed line). b Autocorrelation functions of different sampling processes. The processes with momentum (β = 1) exhibit an oscillatory component. c Mean squared error as a function of Δt. The two cases with an oscillatory component (β = 1) have significantly faster convergence to the true mean (gray line) than those without (β = 0). Fractional Lévy dynamics do not affect the convergence speed in a significant way.
We next elucidate that FNS possesses the powerful computational property of sampling multimodal distributions with far-away modes. For this purpose, we set the target distribution as in Eq. (3) to be a bimodal Gaussian mixture
Without loss of generality, we set s* = 2.5 and σ = 0.32. For clarity and simplicity, here we consider a bimodal distribution, but the results can be generalized to any multimodal distributions. Note that the modal separation Δs* is significantly larger than the modal width σ, implying vanishingly small probability density between the modes; in other words, the modes are separated by a high potential barrier, (U(x)=-log pi (x)), which is extremely difficult to penetrate. For the two fractional cases (α = 1.2), despite the impenetrable energy barrier between the two modes located at ± s*, the sample path is able to switch intermittently between them due to the inherent long jumps of Lévy motion (Fig. 6a). This property enables efficient sampling of the bimodal distribution, with the sample histogram matching the target distribution. In contrast, in both sampling processes driven by conventional Brownian motion (α = 2), i.e., the Langevin Monte Carlo sampling and the Hamiltonian Monte Carlo (HMC) sampling, the sample paths are completely trapped in one of the modes and are unable to explore the entire state space, thus failing to sample bimodal distributions. Particularly, we stress that while HMC has been previously implemented in neural networks for speeding up sampling processes, it cannot on its own resolve the issue of sampling multimodal distributions with far-apart modes because the sampler has a negligible chance to gather a sufficiently large momentum to overcome high potential barriers60.
a Sampling processes driven by fractional Lévy motion (α = 1.2, top two panels) are able to alternate between two modes rapidly (modal separation Δs* = 5), rather than being stuck at one mode as their non-fractional counterparts (α = 2.0, bottom two panels). Right panels: sample histograms and target distribution (solid line). b Mean exit time of the sampling processes driven by fractional Lévy motion grows linearly with modal separation, whereas those driven by Brownian motion grow exponentially; averaged across 24 trials of duration T = 104. Circles and crosses represent results from the mathematical model; solid lines denote the fitted exponential and linear curves.
The Lévy motion can fundamentally improve the mixing ability of the sampling process, that is, the ability to traverse through the probability landscape including the low-probability regions between the two modes. To quantify this mixing property, we calculate the exit time τexit (see “Methods”), which is the duration the random walker remains in one of the modes, and examine how the mean exit time changes with the modal separation Δs*. As shown in Fig. 6b, the dependence of the mean exit time on the modal separation has striking differences between the cases with and without fractional Lévy dynamics. For the two cases driven by fractional Lévy motions, the mean exit time increases linearly with the modal separation; in contrast, for the conventional cases driven by Brownian motion (Langevin sampling and Hamiltonian sampling), either with or without the oscillatory component, the mean exit time blows up exponentially as Δs* increases. Note that these results are achieved with all the other parameters such as the strength of the noise and the shape of the probability landscape fixed. Therefore, the mixing ability can be fundamentally attributed to the fractional Lévy motion ({L}_{t}^{alpha }). The linear dependence of the mean exit time on modal separation is the hallmark of FNS, indicating that it can effectively sample multimodal distributions regardless of variations in modal separation.
As illustrated in the mathematical model of FNS, the drift term b(x) determines the sampled distribution according to Eq. (3). To demonstrate that the spiking neural circuit performs non-trivial probabilistic computations as in the mathematical model, it is necessary to show how the drift term is related to the feedforward input (for encoding sensory evidence) and recurrent synaptic connections (for encoding prior). Due to the mathematical intractability of the spiking neural circuit model, we employ a neural field model (Eq. (27)) to derive the relationships between b(x) and parameters of the feedforward input and the recurrent synaptic weights (see “Methods” for the neural field model and Supplementary Information for details of the mathematical derivations); we then numerically validate these relationships in our spiking neural circuit model, thus providing insights into the probabilistic computations implemented in the spiking neural circuit.
Figure 7a (upper panel) shows b(x) when the neural field model receives a bell-shaped feedforward input (Eq. S5) centered at s1 = 0 with varying contrast levels c1. The drift is positive when x < s1 = 0 and is negative when x > s1 = 0, indicating that the localized pattern experiences a drift toward s1 = 0 and subsequently samples from a probability distribution centered at s1. Further analysis shows that the drift magnitude is proportional to stimulus contrast c1 (Eq. (28)). Using b(x), we then analytically obtain an explicit solution to the resulting sampled distribution (Fig. 7a, lower panel), whose variance is inversely proportional to c1 (Eq. S20). In addition, our analysis shows that a bell-shaped perturbation centered at s0 = 0 to the recurrent synaptic weights (Eq. S6) can induce a drift b(x) with a similar shape around s0 (Fig. 7b, upper panel); the magnitude of the drift is proportional to the strength of synaptic perturbation c0 (Eq. (28)). For this case, we also derive the corresponding sampled probability distribution (Fig. 7b, lower panel) and find that its variance is inversely proportional to c0 (Eq. S21); note that this probability distribution caused by changes in synaptic weights is the prior distribution.
a Drift term (upper panel) and sampled distribution (lower panel) when the neural field model receives a feedforward input with varying contrast levels c1. b Drift term (upper panel) and sampled distribution (lower panel) when the neural field model receives synaptic weight perturbation with varying strengths c0.
We next calculate the drift b(x) in the spiking neural circuit model from the CoM trajectory of the spiking pattern. Since the increments of the CoM are equal to the sum of a deterministic drift and random noise according to the mathematical model (Eq. (2)), we can obtain b(x) by averaging the increments of the CoM trajectory over time and across trials to remove the noise. We then divide the two-dimensional space into bins and calculate the drift as a function of the feature coordinate x. We find that when the neural circuit receives a feedforward input as in Eq. (10), the drift b(x) is positive for x < s1 = 0 and negative for x > s1 = 0, resulting in a sampled distribution centered at s1 (Fig. S2a). A bell-shaped perturbation to the recurrent synaptic weights (Eq. (12)) also yields a similar outcome (Fig. S2b), thus properly embedding the prior to the circuit. These results are qualitatively consistent with the analytical results based on the neural field model, indicating that the drift term is determined by the feedforward input and recurrent synaptic connectivity.
We next validate the powerful mixing property for sampling multimodal distributions in our spiking circuit implementation of FNS. For this purpose, we change the spatial profile of the feedforward input to be the superposition of two Gaussian functions centered at s1 and s2 with contrast levels c1 and c2, respectively (Eq. (11) in “Methods”); for simplicity, here we use a bimodal input as an example, but multimodal distributions can be similarly sampled in the circuit. Such multimodal inputs can be thought as representing sensory responses from multiple sensory channels, as in multisensory perception tasks61. It may also arise from a single stimulus with alternative interpretations such as the spatial orientation of a Necker cube. In both cases, the true state s* of the latent stimulus feature could be either s1 or s2.
Figure 8a shows that the localized pattern wanders around the location of one of the modes for a short while and then jumps to the other one, thus alternately sampling these modes in a similar fashion as in the mathematical model. Consequently, the sampling process leads to bimodal sample histograms whose modes are separated by low-probability regions (right panel in Fig. 8a). To calculate the exit time for the localized activity pattern performing sampling, we assign the spatial coordinates of the CoM trajectory at each time moment to one of the stimulus peaks depending on which peak location it is closer to. The exit time is then defined as the duration the trajectory spends in one region before it switches to the other. We investigate how the exit time changes as the modal separation Δs increases. As shown in Fig. 8b, the mean exit time ranges from 30 to 60 ms, indicating that the alternating sampling between the two modes occurs rapidly. The mean exit time increases linearly for ({{Delta }}s , < , frac{3}{4}pi), before it saturates due to the periodic boundary condition of the circuit model. This result indicates that FNS implemented in the neural circuit exhibits the salient feature as predicted in the mathematical model; that is, the mean exit time of the activity pattern increases linearly with the modal separation.
a Time series of the x– and y-coordinates of the sample path when c1 = c2 = 1. Samples (blue dots) are concentrated around the stimuli (red solid line) and switch rapidly between them without being trapped. Right panel: contour plot of the log sample probability density with a segment of the sample path overlaid on top. b Mean exit time (circle) grows linearly with the modal separation (with saturation due to periodic boundary conditions). c Mean exit time for varying c1 with c2 fixed (c2 = 1). d Mean exit time decreases as both contrast levels are kept the same and increased simultaneously. Error bars in b–d show the standard deviation of the exit time within trials. e Exit time (perceptual dominance duration) follows a right-skewed distribution better fitted with a Burr distribution (red line) with a more slowly decaying tail than a gamma distribution (blue line).
We next demonstrate the effectiveness of our circuit implementation of FNS for different contrast levels by systematically varying c1 while keeping c2 = 1 fixed. Rather than being trapped in the spot with a higher contrast level, the activity pattern is still able to dynamically switch between the two modes intermittently. By calculating the exit time, we find that when the contrast level of one of the stimuli becomes stronger, the mean exit time from this stimulus location increases monotonically while that from the other stimulus location decreases (Fig. 8c).
In summary, these results indicate that FNS can sample multimodal probability distributions in such a flexible and effective way that it is insensitive to their modal separations and contrast differences, thus retaining the excellent mixing ability. These dependencies of the exit time on modal separations and contrast levels form testable predictions of our FNS theory.
When presented with a stimulus with alternative interpretations such as the Necker cube or conflicting stimuli such as in binocular rivalry, humans may experience two alternating percepts31. Such phenomenon known as perceptual switching can be naturally explained in the framework of FNS. To model a simple case of binocular rivalry in which two conflicting stimuli with opponent colors are presented to two eyes, we add inputs to two neural groups whose preferred hues differ by π, as in existing modeling studies62. Alternations between the responses of the two neural groups (left panel, Fig. 8a) are interpreted as switching between different percepts. The exit time from each neural group calculated above is equivalent to the dominance duration of each percept in the context of perceptual switching. The result with varying contrast level c1 and fixed c2 (c2 = 1), as shown in Fig. 8c, is thus consistent with one of the key properties of perceptual switching of binocular rivalry31. When the contrast level of one stimulus is increased from zero with the other fixed, it primarily decreases the mean dominance duration of the stronger percept while increasing that of the weaker one to a lesser extent and equi-dominance is reached when the contrast levels of both stimuli are equal (c1 = c2). Further increasing the contrast level of the stimulus primarily increases the dominance duration of the stronger percept while decreasing that of the weaker one to a lesser extent31. We also increase the contrast levels of the two stimuli simultaneously and find that the mean dominance duration of both percepts decreases (Fig. 8d), which is another key property of perceptual switching31.
In addition, dominance durations in our model follow a right-skewed distribution (Fig. 8e), consistent with psychophysical studies of perceptual switching. Although the specific forms of this type of right-skewed distribution have been found to vary across different experimental and modeling studies, which are still a matter of debate31, we find that the dominance durations in our model are best fitted by a right-skewed distribution called Burr distribution rather than a gamma distribution (see “Methods” for the fitted parameters of these distributions). The key difference is that the former features a slowly decaying power-law tail whereas the latter has a faster decaying exponential tail. It is interesting to note that such a slowly decaying tail in the distribution of perceptual dominance duration has been previously found in experiments63, albeit not explicitly stated. Our results thus propose another right-skewed distribution for capturing dominance duration, which can be tested in future experimental studies.
As in our model, previous noise-driven attractor models have successfully reproduced the dependence of perceptual switching on stimulus contrast and the distribution of dominance duration through a combination of adaptation and externally added noise64. In our model, however, intrinsically emergent fluctuations in the form of fractional Lévy motion are the dynamical origin of perceptual switching. To demonstrate this, in the transition regime of the circuit model without adaption as described above (see Fig. 4), we apply the same feedforward input described by Eq. (11) and find that the localized activity pattern exhibits rapid switching between the two neural groups (Fig. S3). This result thus indicates that the fractional Lévy motion underlies perceptual switching. It is important to note that in our model, the movement of the activity pattern in space has large fluctuation across multiple scales, as evidenced by the heavy-tailed, power-law distribution of its movement step sizes; however, switching behaviors between different attractors possess characteristic scales in existing modeling studies. In ref. 15, it has been shown that noise-driven activity states in a neural network model can be used to perform MCMC sampling and to model perceptual switching. In this study, the effect of modal separation on perceptual switching was not investigated; however, as demonstrated in the mathematical model (Eq. (2)), at the algorithmic level, the MCMC sampler with Langevin dynamics is unable to traverse through low-probability regions, particularly when multimodal distributions have far-apart modes. In contrast, the FNS theory suggests that Lévy motion with large fluctuations emerging from the spiking neural circuit can serve as a robust mechanism for explaining perceptual switching, as the FNS-based sampling dynamics are insensitive to modal separation.
We next illustrate the properties of FNS-based probabilistic inference particularly through accounting for the key features of human perceptual inference in a motion direction task32. In this experiment, subjects were presented with visual stimuli consisting of moving dots, some of which move in the same direction while the rest move in random directions; the fraction of dots moving in the same direction is called coherence. It has been shown that the mean and variance of perceptual estimates match those predicted by the basic Bayesian observer model. However, the distribution of the perceptual estimates was found to be bimodal32; this observation was at odds with the basic Bayesian observer model which gave rise to a unimodal posterior. We now demonstrate that FNS-based inference provides a neural mechanism for resolving this puzzle and further makes new predictions about perceptual inference.
To proceed, we first establish a connection between the experimental setup and our model by interpreting the x-component of the 2D feature space as motion direction. The x-component of the CoM of the localized pattern at a particular time instant thus represents the perceived motion direction, whereas the y-component is considered as representing a nuisance parameter. We assume that the visual stimulus is represented through a population code as a bell-shaped feedforward input to the spiking neural circuit model (Eq. (10)), such that its center s1 represents the motion direction and its contrast c1 is proportional to the motion coherence, similar to that used in ref. 32. With this understanding, we compare the FNS-based inference results with the experimental results by carrying out the similar procedures as in the original experiment. In the experiment, the subjects were first trained with a block of visual stimuli to learn a prior distribution, and then were presented with a visual stimulus for a brief period of 300 ms before the subject was required to report the motion direction. Correspondingly, in our spiking neural circuit model we first embed a prior by applying a bell-shaped perturbation to the synaptic weights centered at s0 with amplitude c0 according to Eqs. (12) and (13) (see “Methods”). As illustrated above, these synaptic changes can be used to properly embed a prior to the recurrent circuit (Fig. 7). We then add a unimodal feedforward input with contrast c1 centered at s1 (Eq. (10)), which encodes the sensory evidence about the motion direction of the visual stimulus. The sampler (i.e., the localized activity pattern) is able to switch freely between the prior and the newly added sensory evidence within the short stimulus presentation period. The instantaneous position of the activity pattern at the end of the stimulus presentation is then interpreted as the perceptual estimate.
We then use our model to explain the three sets of experiments in ref. 32. The first set of experiments involves fixing the coherence of the random moving dots and the prior while varying the motion direction relative to the prior mean. Correspondingly, in our neural circuit model, we fix the contrast level c1 = 0.5 in Eq. (10) and the strength of synaptic perturbation c0 = 0.002. Without loss of generality, we fix the prior mean s0 = (0, 0) and systematically vary the center s1 = (s, 0) of the feedforward input with s varying from −π to π. As shown in Fig. 9a (left panel), the estimate distribution is characterized by two peaks: one corresponding to the prior centered at the origin and another corresponding to the sensory evidence with varying motion directions. Notably, the similar feature of bimodality versus modal separation has been observed in the experiments32 (see Fig. 4 of their paper). As illustrated above, the bimodality in our circuit model arises as the spiking activity pattern intermittently alternates between the two modes (Fig. 8a). We further quantify the FNS-based probabilistic inference by calculating the mean and variance of perceptual estimates across trials. As the center of the feedforward input (thus the modal separation relative to the prior at the origin) increases from −π to π, the mean of the bimodal perceptual estimate distribution also increases from negative to positive values in an approximately linear fashion (Fig. 9a, top right panel); this relationship is consistent with the experimental results32 (see Fig. 3a of their paper). The variance of the perceptual estimate is approximately a quadratic function of modal separation (Fig. 9a, bottom right panel). We find that the linear and quadratic dependences of the mean and variance of perceptual estimates on the modal separation are also consistent with the analytical results obtained from the neural field model (see Eqs. S23–S24 in Supplementary Information).
Perceptual estimate histogram, mean and standard deviation for: a varying modal separations between the stimulus (red bar) and the prior (black bar); b for varying stimulus contrast and fixed prior; c for varying prior strength (synaptic perturbation) and fixed stimulus contrast. Shades indicate the standard deviation across 20 trials. d, e Statistics of estimate distribution based on reanalysis of published experimental data. d Dependence of perceptual estimate variance (dots) on the modal separation can be approximately fitted by quadratic functions (solid curves), excluding outliers due to periodic boundary conditions. Stimulus coherence is fixed at 6%. e Standard deviation of the perceptual estimates decreases with stimulus coherence (solid lines), except for low coherence levels. Dashed lines denote extrapolations based on the theoretical argument that the perceptual estimate distribution should be equal to the prior when the coherence is zero. Modal separation is fixed at 20°.
To test the predictions from our inference results, we re-analyze the published data32 and find that the variance of the perceptual estimates also matches a quadratic curve when the prior standard deviation is large (80 degrees), as shown in Fig. 9d (red dots). However, for smaller prior standard deviations (Fig. 9d, green and orange dots), whether the experimental data matches our prediction remains inconclusive due to the limited data range32. Specifically, since the perceptual stimuli used in the experiments were pooled from the same dataset with which the prior was trained, its range was limited by the prior standard deviation (e.g., for a prior std of 10 degrees, the range of the presented stimulus is only about ±20 degrees). Future experiments could test our predictions by extending the range of the perceptual stimuli even when the prior standard deviation is small.
The second set of experiments involves varying the motion coherence while fixing the prior uncertainty and the distance of motion direction relative to the prior mean32. In our neural circuit model, this corresponds to varying the contrast level of the feedforward input c1 while fixing the strength of the synaptic perturbation c0 = 0.002 and the modal separation between them. The sample distributions for this case are shown in Fig. 9b (left panel), likewise exhibiting bimodality. As the contrast level increases, the height of the mode corresponding to sensory evidence increases whereas the height of the prior decreases. The mean and variance of perceptual estimates are also largely consistent with the experimental results: as the contrast level of the stimulus increases from zero, the mean of the bimodal perceptual estimate distribution gradually shifts from the prior at s0 toward the sensory evidence at s1 (Fig. 9b, top right panel). Meanwhile, the perceptual estimate variance decreases with the contrast level, but with the exception at low contrast levels (Fig. 9b, bottom right panel).
This finding that the variance of perceptual estimates varies non-monotonically with the stimulus coherence therefore provides a testable prediction for future experimental studies. By re-analyzing the published data32, we find that the perceptual estimate standard deviation decreases with the stimulus coherence for a number of coherence values (Fig. 9e). Experimental data for coherence smaller than 5% is unavailable so whether monotonicity is violated cannot be directly verified. However, if we impose the theoretical argument that the perceptual estimates should simply be samples from the prior distribution when the coherence is zero, we can then extrapolate the estimate standard deviation with the prior standard deviation. By presenting all data points together, we find that the monotonicity is indeed violated (Fig. 9e). To directly test our prediction, future experiments should more closely examine changes in the variance of perceptual estimates when the stimulus contrast levels are low.
The third set of experiments involves varying the uncertainty of the prior while fixing the coherence of the stimulus and the motion direction32. In the neural circuit model, this corresponds to varying the strength of the heterogeneous synaptic perturbation c0 while fixing the contrast of the feedforward input c1 = 0.5 and the modal separation. As the synaptic perturbation becomes stronger, the height of the prior increases and the height of the mode corresponding to sensory evidence decreases (Fig. 9c, left panel). The perceptual estimate mean gradually shifts from the sensory evidence at s1 to the prior s0 (Fig. 9c, top right panel). The perceptual estimate variance also depends non-monotonically on the strength of synaptic perturbation (Fig. 9c, bottom right panel). This observation cannot be verified in the experimental data due to the limited data size and thus forms a testable prediction for future experiments.
In this study, we have presented a theory (i.e., FNS) of probabilistic neural computations through the illustration of its neural circuit implementation as well as the normative formulation of its computational properties. By extending existing models that are mainly based on temporally variable or fluctuating dynamics13,14, FNS exploits rich, complex fluctuations of neural population activity both in time and in space for efficiently performing sampling-based probabilistic computations; FNS thus offers an approach to addressing the long-standing challenge of sampling and representing multimodal distributions. Our probabilistic neural computation theory provides a unified account of a variety of findings on neural dynamics at the individual neuron and circuit levels as well as on perceptual phenomena such as perceptual switching and visual perception inference, thus establishing a framework for understanding neurophysiological and computational mechanisms of brain function.
Our FNS theory provides a new perspective on the role of complex spatiotemporal neural dynamics that exhibit large fluctuations across multiple scales, which are of a non-Gaussian (i.e., heavy-tailed) nature. Particularly, FNS-based probabilistic computations harness the power of the fractional Lévy motion of population activity patterns (i.e., neural ensembles); these patterns hover around one location for a while and then move or switch to another location in an intermittent manner, with their movement step sizes following heavy-tailed, power-law distributions. Such fractional motions in space give rise to irregular propagation trajectories and speeds with large variability. Propagating activity patterns have been widely observed at the circuit and the whole brain levels22,23; notably, localized gamma activity patterns with Lévy motion have been found in the MT area of marmoset monkeys45. We have repeated the analysis on the broadband LFP from ref. 45 (see Supplementary Information) and found that localized activity patterns of such broadband activity also show Lévy motion (Fig. S4). It has been shown that hippocampal sharp wave ripples exhibit movements with clusters of short step sizes that are intermittently interspersed by long jumps46. Nevertheless, to directly test our modeling prediction of propagating spiking patterns with Lévy motion, future studies need to focus on massive individual-neuron recordings and to analyze spiking patterns in the same way as done in our modeling study.
Another key property of sampling dynamics of FNS is that the autocorrelation function of the sample path exhibits an oscillatory component. Such an oscillatory component induces negative-going lobes in the autocorrelation; consequently, this would result in independent samples occurring in a much shorter time, thus playing an essential role in increasing the effective sampling speed as in the Hamiltonian Monte Carlo method57,58. In a previous modeling study58, such oscillatory activity arises from mapping Hamiltonian dynamics to the activity of excitatory and inhibitory neurons and its peak frequency is around 40 Hz (i.e., gamma oscillation). Importantly, it has been demonstrated that such gamma oscillations can naturally emerge in a circuit model trained for optimally performing sampling-based probabilistic inference65. In our model, as illustrated in the autocorrelation analysis of the sampling process, the temporal oscillation is in the theta range (3–8 Hz). We have found that the spike frequency adaptation is essential for the genesis of the theta oscillatory component, in accordance with other modeling studies56. However, different from oscillatory activity with clock-like periodicity in these models, neural oscillations in our model exhibit great variability and are accompanied by heavy-tailed, 1/f-like activity, as shown in the power spectrum of the firing rates of neural population activity. Notably, such temporal fluctuations with a 1/f component have been widely observed in neural population activity as recorded by LFP, EEG, and MEG during both spontaneous activity and task conditions20. Traditionally, such 1/f-like activity has been deemed unimportant and often removed from analyses to emphasize oscillatory components. In our theory, however, it is an integral part of our circuit implementation of FNS.
It has been found that theta oscillations are a hallmark feature of environmental sampling having now been linked to spatial attention sampling 30,52, eye movements in primates66, and whisking in rodents49. In some of these environmental sampling tasks, such as eye movement-based sampling of natural scenes, the existence of Lévy-like motion has been reported67. These observations lead us to predict that our FNS-based probabilistic computation might be the underlying computational mechanism of these cognitive functions; indeed, recently we have successfully applied FNS dynamics to explain key neural and behavioral effects of visual attention51. Furthermore, it is interesting to note that the phase of theta oscillations often modulates the amplitude of gamma fluctuations68. Such theta-gamma coupling suggests that it is important to explore whether and how each theta sampling cycle can be implemented through gamma activity, potentially unifying our FNS approach with the gamma dynamics-based sampling approaches58,65 and gaining a comprehensive understanding of probabilistic computations in neural circuits.
As formulated in the mathematical model with fractional order derivative and further illustrated in the spiking neural circuit, FNS exploiting the complex spatiotemporal dynamics mentioned above possesses some profound computational advantages. The sampling processes of FNS are quicker than those of the other methods such as the classical Langevin sampling method that has been implemented in stochastic recurrent neural networks. Importantly, FNS enables the efficient sampling of multimodal distributions even with far-away modes, due to the long jumps inherent to fractional movements of neural activity patterns. Sampling multimodal distributions is crucial for probabilistic computations, which, however, would not be possible in existing models, because they rely on Gaussian dynamics (i.e., Brownian motion) for implementing classical Langevin Monte Carlo sampling methods used in machine learning69. These methods are notorious for becoming trapped in a single mode, and lack the capacity to efficiently traverse through complex probability landscapes. Based on both the circuit and mathematical models, our theory of FNS makes a core prediction; that is, the mean exit time (i.e., the average time the sample path takes to leave one of the modes of a bimodal distribution) depends linearly on the modal separation. This prediction is falsifiable and hence represents a strong test of our FNS theory.
Our FNS can be implemented in biophysically realistic, two-dimensional spiking neural circuits incorporating two well-established properties of the cortex. One property is the dependence of synaptic connection probability on distance36,37 and another is the balance between excitatory and inhibitory inputs, with the former closely tracked by the latter38. Most previous models of sampling-based representations14,15,58,59, however, have focused on either normative probabilistic models or theories without specifying their underlying neural circuit mechanisms14,15, or on abstract neural circuits models 58, some of which have no separation of excitatory and inhibitory neurons thus violating Dale’s principle59. Networks with one-dimensional structure (i.e., a ring structure) have been trained to produce a fast sampling-based inference exploiting cortical-like temporal dynamics65. Since these networks do not intrinsically produce variable neural dynamics, an external source of fluctuations is required. In contrast, in our circuit model, spatiotemporal activity patterns with large fluctuations are intrinsically generated and the 2D spatial structure supports the emergence of such fluctuations. This scenario is similar to other complex physical systems where such a 2D spatial extension is important for the emergence of spatiotemporal patterns with complex dynamics70. In our model, the complex spatiotemporal patterns with fractional motions only emerge when the circuit works near the transition between different activity states; these findings are consistent with recent studies proposing that complex cortical dynamics could be better understood in the dynamical regime close to the transition between different cortical states (i.e., asynchronous and synchronous/coherent states)71,72,73, but go beyond them by revealing the fundamental functional role of complex spatiotemporal cortical dynamics in probabilistic neural computations.
The circuit mechanism of FNS also has important implications for constructing efficient artificial neural networks in machine learning. Large-scale artificial neural networks such as the Boltzmann machine74 and deep belief networks75 often perform complex computations through employing probabilistic sampling. Most of these artificial neural networks perform probabilistic sampling by using the classical MCMC driven by Gaussian noise, so they suffer from the similar problem as faced by the existing models of neural probabilistic sampling, in that these methods lack the ability to jump across low-probability regions and to traverse through the probability landscape (also known as the ‘mixing’ ability). The problem of mixing becomes even more pronounced when dealing with larger complex datasets. The common remedy for these methods relies on some variants of simulated tempering76, which change the temperature parameter in order to globally flatten the solution landscapes during the sampling process. These tempering methods, however, come with a cost of their own, because they require extra computations and parameter-tuning that assume knowledge about the global state of the artificial neural networks. However, our circuit-based FNS mechanism indicates that powerful mixing abilities for representing multimodal distributions can emerge from the circuits that are essentially locally coupled, and importantly this happens in a fundamentally autonomous manner without tuning any global parameters during the sampling process. Motivated by these properties of the circuit-implementation of FNS, we thus suggest that future large-scale, information processing neural network models may benefit from our circuit mechanism of FNS, if they are designed to exploit fractional Lévy-like diffusion and oscillations for sampling complex, high dimensional probability landscapes.
As we have demonstrated, the FNS-based probabilistic sampling can implement perceptual inference that explains both perceptual estimate statistics and the bimodality of perceptual estimate distributions as found in a recent study of motion direction estimation task32. Regarding probabilistic inference, our model makes two main predictions: first, the perceptual estimate variance is approximately a quadratic function of the modal separation between the sensory evidence and the prior; second, the perceptual estimate variance decreases with stimulus contrast but this trend is reversed when the contrast level is small. These model predictions are consistent with our reanalysis of the experimental data32, except for the cases when the stimulus contrast level is low; a conclusive test of these predictions will require more specific manipulation of stimuli to cover the low contrast range. Previously, a binary switching model with Gaussian noise was used to explain the bimodal property of perceptual estimates32. In our model, due to the non-equilibrium, fractional nature of Lévy motion, the activity patterns exhibit short-step clusters that are interspersed by long jumps, thus giving rise to clustering and switching-like behaviors and the resulting bimodality of the estimate probability distributions. Rather than being simplistic and binary, the switching-like behavior in our FNS model happens across multiple scales with large fluctuations, as evidenced by step sizes with heavy-tailed, power-law distributions. Switching-like dynamics with large fluctuations have indeed been found in the inferotemporal area of monkey during sensory processing33, in the orbitofrontal cortex34, and in the lateral intraparietal area35 during decision making. Our FNS mechanism of probability inference might therefore be generally applicable to understanding these key brain functions.
The spatially extended spiking neural circuit model consists of NE excitatory neurons and NI inhibitory neurons embedded in a two-dimensional feature space with periodic boundary conditions40. Each neuron i is assigned to spatial coordinates si = (xi, yi) representing the preferred stimulus features (orientation and color) of that neuron, with both xi and yi ranging from −π to π. We consider NE = 63 × 63 = 3969 excitatory neurons on a square grid and NI = 1000 inhibitory neurons with uniformly random locations. The ratio between NE and NI is around 4. The Euclidean distance is calculated between each pair of neurons i from population α and j from population β, where α and β are either excitatory (E) or inhibitory (I). The connection probability between neurons is proportional to the distance-dependent factor ({{{Omega }}}_{ij}^{alpha beta }={e}^{-{D}_{ij}^{alpha beta }/{tau }_{D}^{alpha beta }}). For the excitatory connections the spatial scales are ({tau }_{D}^{{{{{{{{rm{EE}}}}}}}}}=8) and ({tau }_{D}^{{{{{{{{rm{IE}}}}}}}}}=10) grid units, whereas for the inhibitory connections the spatial scales are ({tau }_{D}^{{{{{{{{rm{II}}}}}}}}}={tau }_{D}^{{{{{{{{rm{EI}}}}}}}}}=20) grid units.
The subthreshold membrane potential ({V}_{i}^{alpha }) of neuron i in population α follows
where the membrane capacitance C = 0.25 nF, the leak conductance gL = 16.7 nS, and VL = −70 mV is the reversal potential for the leak current. ({I}_{i,{{{{{{{rm{K}}}}}}}}}^{alpha }(t)) is the potassium current, ({I}_{i,{{{{{{{rm{rec}}}}}}}}}^{alpha }(t)) is the recurrent synaptic current received by the neuron and ({I}_{i,{{{{{{{rm{ext}}}}}}}}}^{alpha }(t)) is the external current. When the membrane potential reaches the threshold Vth = − 50 mV, a spike is emitted and the membrane potential is reset to the potential Vrt = −60 mV for an absolute refractory period τf = 4 ms. The potassium current is given by ({I}_{i,{{{{{{{rm{K}}}}}}}}}^{alpha }(t)=-{g}_{i,{{{{{{{rm{K}}}}}}}}}^{alpha }(t)({V}_{i}^{alpha }(t)-{V}_{{{{{{{{rm{K}}}}}}}}})), where ({g}_{i,{{{{{{{rm{K}}}}}}}}}^{alpha }(t)) is the active potassium conductance and VK = −85 mV. The dynamics of the potassium conductance are described by
where ({t}_{j,k}^{alpha }) is the time of the kth spike emitted by neuron i from population α, ΔgK = 10 nS and τK = 80 ms. Because spike frequency adaptation has been primarily observed in cortical pyramidal neurons, we only include such adaptation for excitatory neurons in our model. The recurrent synaptic current ({I}_{i,{{{{{{{rm{rec}}}}}}}}}^{alpha }(t)) in Eq. (5) is
where ({g}_{i}^{alpha beta }(t)) is the conductance of the recurrent current from the presynaptic population β. The excitatory and inhibitory reversal potentials are ({V}_{{{{{{{{rm{rev}}}}}}}}}^{{{{{{{{rm{E}}}}}}}}}=0) mV and ({V}_{{{{{{{{rm{rev}}}}}}}}}^{{{{{{{{rm{I}}}}}}}}}=-80) mV, respectively. The conductance ({g}_{i}^{alpha beta }(t)) is given by
where ({a}_{ij}^{alpha beta }) and ({J}_{ij}^{alpha beta }) represent the coupling topology and the connection strength, respectively, as detailed in the original paper40. Specifically, the synaptic connection probability between neurons decreases in an exponential manner as distance between neurons increases, consistent with experimental findings36,37. The non-dimensional gating ({s}_{ij}^{alpha beta }(t)) describes the synaptic dynamics
where ({tau }_{{{{{{{{rm{d}}}}}}}}}^{beta }) and ({tau }_{{{{{{{{rm{r}}}}}}}}}^{beta }) are the decay and rise time constants, respectively, ({t}_{j,k}^{beta }) is the time point of the kth spike of neuron j from population β, and ({d}_{ij}^{alpha beta }) is the conduction delay drawn from a uniform distribution between 0 and 4 ms.
In our model, we consider an essential neurophysiological feature of local cortical circuits, that is, the excitatory post-synaptic currents are proportional to the inhibitory ones, with a homogeneous ratio across all excitatory neurons, as found in mouse primary visual cortex layer 2/353. To model this in our heterogeneous circuit, we consider the I–E ratio ({xi }_{i}=mathop{sum }nolimits_{k=1}^{{K}_{i,{{{{{{{rm{in}}}}}}}}}^{{{{{{{{rm{EI}}}}}}}}}}{J}_{jk}^{{{{{{{{rm{EI}}}}}}}}}/{sum }_{j}{J}_{ij}^{{{{{{{{rm{EE}}}}}}}}}), where ({K}_{i,{{{{{{{rm{in}}}}}}}}}^{{{{{{{{rm{EI}}}}}}}}}) denotes the number of connections (in-degree) received by excitatory neuron i from the inhibitory population and the connection strength values ({J}_{ij}^{{{{{{{{rm{EE}}}}}}}}}) are determined by the reverse pooling method40. To equalize the I–E ratio ξi across the neurons to a desired network-wide ratio, that is, 〈ξi〉 = ξ, the ({J}_{ij}^{{{{{{{{rm{EI}}}}}}}}}) for neuron i are sampled from a Gaussian distribution with a mean equal to (xi {sum }_{j}{J}_{ij}^{{{{{{{{rm{EE}}}}}}}}}/{K}_{i,{{{{{{{rm{in}}}}}}}}}^{{{{{{{{rm{EI}}}}}}}}}) and a standard deviation that is 25% of the mean. The I–E ratio ξ is varied as a system parameter for demonstrating the emergence of localized activity patterns with Lévy motion (Fig. 4) but is otherwise fixed at the value ξc = 3.4.
It has been shown previously that the circuit model exhibits a rich repertoire of dynamical activity states, ranging from asynchronous to localized and global propagating wave states40. Particularly, it has shown that around the phase transition between the asynchronous and localized propagating wave states, our circuit model exhibits nonlinear response properties different from the classical balanced state and can quantitatively reproduce a variety of major empirical findings regarding neural spatiotemporal dynamics. The phase transition was characterized by calculating the susceptibility and the branching parameter of spikes40. In the present study, we only change one parameter, the I–E ratio, to study the spatiotemporal properties (i.e., Lévy motion in space and theta oscillations in time) of the dynamical activity pattern emerging in this transition regime and their fundamental roles in probabilistic computations.
To study how the activity pattern samples non-trivial distributions, external stimuli are applied in the form of independent Poisson spike trains with specific firing rates. First, for sampling unimodal distributions, we apply an external feedforward input to each neuron i in the recurrent population as independent Poisson spikes whose firing rate is equal to
where σ = 0.6 is the width of the feedforward input, and c ≥ 0 is the contrast. When c = 0, the network receives the baseline background input with a uniform rate r0 = 0.85 kHz. The true value of the stimulus feature is fixed at s* = (0, 0) without loss of generality. Second, for sampling bimodal distribution, we apply an external feedforward input with firing rate equal to
The baseline firing rate is reduced to r0 = 0.5 kHz so that spontaneous activity (corresponding to when c1 = c2 = 0) is suppressed, ensuring a low-probability region between the modes of the sampled bimodal distribution. The true value of the stimulus feature s* is unknown by the observer and could be either of the alternative interpretations s1 or s2. We investigate the performance of our network in sampling bimodal distributions by varying the stimulus contrast levels c1 and c2 and the modal separation Δs = ∥s1 − s2∥.
To embed prior in the recurrent synaptic connectivity, we apply a spatially heterogeneous perturbation to the excitatory-to-excitatory synaptic weight
where si = (xi, yi) is the coordinates of neuron i in the feature space, c0 is the magnitude of the weight perturbation, and ϕ is a two-dimensional bell-shaped function
centered at s0 = (x0, y0) with the width parameter d. In this study, we fix s0 = (0, 0) and d = 1.3.
Our spatially extended network exhibits localized spiking activity patterns with complex spatiotemporal dynamics. To characterize the dynamics of the pattern, we first track its center of mass (CoM), ({hat{{{{{{{{bf{s}}}}}}}}}}_{t}=({hat{x}}_{t},{hat{y}}_{t})) based on the population vector of excitatory neurons77
where xj ∈ [−π, π) is the x-coordinate of the jth excitatory neuron, nj is the spike count over a small time window [t − τ, t) (here we set τ = 15 ms), i is the imaginary unit, and Arg is the principal complex argument in [−π, π). The y-coordinate of the CoM ({hat{y}}_{t}) is defined similarly. From a population coding point of view, Eq. (14) is also known as a complex estimator, which is equivalent to fitting a cosine function to the spike counts in the least-square sense78. In our study, the trajectory of the CoM of the activity pattern is the sample path of FNS.
Based on the trajectories of CoM, we next calculate the increments and the MSD. Specifically, the raw coordinates ({hat{{{{{{{{bf{s}}}}}}}}}}_{t}) over the periodic space are first unwrapped to linear coordinates in order to accommodate distance calculations across the periodic boundary. The mean-squared displacement of the CoM is then ({{{{{{{rm{MSD}}}}}}}}({{Delta }}t)=langle parallel {hat{{{{{{{{bf{s}}}}}}}}}}_{t+{{Delta }}t}-{hat{{{{{{{{bf{s}}}}}}}}}}_{t}{parallel }^{2}rangle), where Δt is the time lag. For fitting the increment distribution, we set ΔT = 15 ms to ensure no temporal correlation is artificially introduced due to overlaps in the spike count time window. The increments are then fitted to a symmetric Lévy stable distribution ({{{{{{{mathcal{S}}}}}}}}alpha {{{{{{{mathcal{S}}}}}}}}(alpha,gamma )) with the tail exponent α ∈ (0, 2] and the scale parameter γ > 0, as defined by Eq. (15). The parameters of the distribution are estimated with maximum likelihood. The MSD is averaged across 100 independent realizations of networks, each collecting 50 s of samples, whereas the increment histogram is produced with samples collected over 10 s, repeated across 5 random realizations of the network.
To characterize the state transition from the asynchronous to propagating wave state, we calculate local pairwise correlations of spike counts. We first divide the two-dimensional network into non-overlapping local patches of size 9 × 9 grid units and then calculate the pairwise spike count correlation within each patch. Finally, the mean local pairwise correlation is obtained by averaging the correlation coefficients of individual patches (49 patches in total). A weak local pairwise correlation indicates asynchronous activity without any structured patterns, whereas a strong local pairwise correlation indicates a spatially coherent wave pattern. The spike count time window is set to 100 ms with a total simulation time of 10 s repeated over 5 independent trials.
To understand the algorithmic nature of FNS implemented by our spiking circuit model, we develop a mathematical model based on stochastic dynamics driven by Lévy motion, rather than Brownian motion as in the classical Langevin MCMC. We consider a class of Lévy motion ({L}_{t}^{alpha }) whose increments follow symmetric Lévy stable distributions, denoted as ({{{{{{{mathcal{S}}}}}}}}alpha {{{{{{{mathcal{S}}}}}}}}(alpha,gamma )), with a tail index α and a scale parameter γ. For clarity and analytical tractability, we restrict our discussion to the one-dimensional case. The probability density of ({{{{{{{mathcal{S}}}}}}}}alpha {{{{{{{mathcal{S}}}}}}}}) can be expressed as28
The probability density exhibits a heavy tail, with power-law asymptote of the form79
for all 0 < α < 2. Note that α = 1 corresponds to the special case of a Cauchy distribution and α = 2 corresponds to a Gaussian distribution. The increment ({{Delta }}{L}_{t}^{alpha }={L}_{t}^{alpha }-{L}_{t^{prime} }^{alpha }) with ({{Delta }}t=t-t^{prime} > , 0) follows ({{{{{{{mathcal{S}}}}}}}}alpha {{{{{{{mathcal{S}}}}}}}}(alpha,{{{Delta }}t}^{frac{1}{alpha }})).
To derive the mathematical model described by Eq. (2), we start with the general formula presented in ref. 80
where ({{{{{{{{bf{z}}}}}}}}}_{{{{{{{{bf{t}}}}}}}}}={({x}_{t},{v}_{t})}^{T}), D is a positive semi-definite matrix describing the fractional diffusion, and Q is a skew-symmetric matrix describing the interplay between position xt and momentum vt. The temporal evolution of the probability density p(z, t) is given by the fractional Fokker-Planck equation,
where the fractional Riesz derivative ({{{{{{{{mathcal{D}}}}}}}}}^{alpha }) is defined as
through the Fourier transform ({{{{{{{mathcal{F}}}}}}}}[f]=int d{{{{{{{bf{z}}}}}}}}{e}^{-i{{{{{{{bf{k}}}}}}}}cdot {{{{{{{bf{z}}}}}}}}}f({{{{{{{bf{z}}}}}}}})). Fractional order derivatives generalize the notion of differentiation to fractional orders and are powerful mathematical tools for describing complex dynamics81; recently, fractional Fokker-Planck equations have been used to explain how non-Gaussian neural dynamics emerge from biologically realistic neural circuits82. It can be shown that the temporal evolution of p(z, t) converges to a stationary distribution (pi ({{{{{{{bf{z}}}}}}}})propto phi ({{{{{{{bf{z}}}}}}}})=exp (-H({{{{{{{bf{z}}}}}}}}))=exp (-U(x)-T(v))), related to the drift term by
To retain both the efficient sampling provided by the Hamiltonian dynamics as well as the powerful mixing ability provided by the fractional dynamics, we set ({{{{{{{bf{D}}}}}}}}=big[begin{array}{ll}gamma &0\ 0&0end{array}big]) and ({{{{{{{bf{Q}}}}}}}}=big[begin{array}{ll}0&-beta \ beta &0end{array}big]) and obtain
Although convergence to the stationary distribution is guaranteed mathematically, numerically simulating this set of equations is challenging since there are no straightforward ways for evaluating the Riesz fractional derivative with two dimensions. To overcome this numerical challenge, we exploit the fact that we are only interested in sampling the target distribution of π(x) instead of the joint stationary distribution of π(x, v) and make two simplifications as follows. First, we approximate the Riesz fractional derivative ({{{{{{{{mathcal{D}}}}}}}}}^{alpha }) with its partial versions ({{{{{{{{mathcal{D}}}}}}}}}_{x}^{alpha }) and ({{{{{{{{mathcal{D}}}}}}}}}_{v}^{alpha }), such that
The partial Riesz fractional derivative can then be effectively evaluated using a fractional centered difference scheme83,84. We find that this approximation in practice provides the correct sample distribution through numerical simulation. Second, by setting (psi (v)=exp (-T(v))) to be a symmetric α-stable distribution SαS(1/α), it holds that
By applying these two simplifications, we obtain the fractional Hamiltonian dynamics for modeling fractional neural sampling (Eq. (2)). As we show numerically in Results, the distribution of the samples generated by Eq. (2) converges to the target distribution π(x). For all simulations we use the Euler-Maruyama scheme with a step size of Δt = 0.001 with γ = 1, and apply a clipping (tilde{b}(x)approx min (b(x),{b}_{max })) with ({b}_{max }=500) to avoid numeric overflow of the Riesz derivative near low-probability regions.
A number of properties about the Riesz derivative are worth noting. First, when α = 2, the Riesz fractional derivative coincides with the ordinary second-order derivative with a change of sign, that is, ({{{{{{{{mathcal{D}}}}}}}}}^{2}=-{partial }_{xx}-{partial }_{vv}). Second, unlike integer-order derivatives, the Riesz derivative cannot be broken down into the sum of the partial derivatives in each of its dimension, that is, ({{{{{{{{mathcal{D}}}}}}}}}^{alpha }, ne , {{{{{{{{mathcal{D}}}}}}}}}_{x}^{alpha }+{{{{{{{{mathcal{D}}}}}}}}}_{v}^{alpha }), as the right-hand side of this equation cannot preserve isotropy in the space spanned by (x, v).
To characterize the ability of FNS for sampling bimodal distributions, we calculate how the mean exit time, which measures the average duration the sampler spends near one of the modes, changes as a function of modal separation. Suppose that the two modes of a bimodal distribution are centered at s1 and s2, then the exit time τexit from the first mode to the second mode can be defined as the duration the sample trajectory remains closer to s1 than to s2
where the ({hat{{{{{{{{bf{s}}}}}}}}}}_{0}) is any point satisfying (parallel {hat{{{{{{{{bf{s}}}}}}}}}}_{{0}_{+}}-{{{{{{{{bf{s}}}}}}}}}_{1}parallel le parallel {hat{{{{{{{{bf{s}}}}}}}}}}_{{0}_{-}}-{{{{{{{{bf{s}}}}}}}}}_{2}parallel) with ({hat{{{{{{{{bf{s}}}}}}}}}}_{{0}_{pm }}) denoting the one-sided limits (mathop{lim }nolimits_{tto pm 0}{hat{{{{{{{{bf{s}}}}}}}}}}_{t}). The mean exit time is then calculated as the average of all exit times in a sample trajectory. This definition applies to both the 1D mathematical model and the 2D neural circuit implementation. For the mathematical model, we use an open boundary condition rather than a periodic boundary condition in order to highlight the impact of modal separation on sampling. For the case of periodic boundary condition, the exit time increases similarly for small modal separations but saturates for large modal separations.
In the context of perceptual bistability, the exit time can be interpreted as perceptual dominance duration which is known to follow a right-skewed distribution. We fit the exit time calculated from the neural circuit model to two candidate distributions using the method of maximum likelihood. The first one is the Burr distribution with probability density function
with parameters found to be λ = 65.0 ms (95% CI 62.7–65.3), c = 8.05 (95% CI 7.58–8.55), and k = 0.528 (95% CI 0.476–0.586). The second one is the gamma distribution with probability density
with parameters found to be k = 10.4 (95% CI 9.9–10.9) and θ = 7.51 ms (95% CI 7.14–7.88). Note that the former features a power-law tail (with an exponent equal to −1 − ck) whereas the latter features a faster decaying exponential tail. Outliers (dominance duration <29 ms) are occasionally produced during the asynchronous phase of the spiking pattern and thus are omitted for this analysis.
To gain theoretical insights into how the drift term and sampled distribution are related to sensory input and synaptic weights, we consider a continuous neural field model85, which captures some features of the spiking circuit model, including the distance-dependent synaptic coupling and a localized activity pattern (bump activity) that moves randomly in the presence of noise. The continuous neural field model with one spatial dimension is described as
where u(x, t) represents the synaptic current, (J(x !,, x^{prime} )) is the synaptic coupling function, ρ is the neural density, I(x) is the feedforward input current, and F is a quadratic neural activation with divisive normalization. We consider the following form of synaptic coupling (J(x !,, x^{prime} )=bar{J}(x-x^{prime} )+{c}_{0}tilde{J}(x !,, x^{prime} ;{s}_{0})), where (bar{J}) is a translation-invariant synaptic coupling function and (tilde{J}) is a heterogeneous synaptic perturbation centered around s0, and c0 is the strength of the perturbation. Similarly, the feedforward input is (I(x)=bar{I}+{c}_{1}tilde{I}(x;{s}_{1})+xi (x),) where (bar{I}) is a constant background input, (tilde{I}(x;{s}_{1})) is a stimulus-dependent feedforward input centered at s1, c1 represents stimulus contrast, and ξ(x) is an additive noise; see Supplementary Information for the detailed definitions of these terms. By applying a projection method, we obtain an explicit analytical expression for the drift term
where a is the width of the homogeneous synaptic coupling, d1 is the width of the feedforward input, d0 is the width of the synaptic perturbation, and r0 is the height of the localized activity. The corresponding sampled distribution is bimodal when s1 and s0 are far apart, in which case we can apply Laplace approximation to cast it into the form of a Gaussian mixture
where g(x; μ, σ2) denotes a Gaussian distribution with mean μ and variance σ2. The variance of each of the modes is inversely proportional to the strength of perturbation c1 or c0, with the constant of proportionality equal to ({kappa }_{1}=frac{gamma {c}_{alpha }tau {({a}^{2}+{d}_{1}^{2})}^{3/2}}{2sqrt{2}{a}^{2}{d}_{1}}), or ({kappa }_{0}=frac{gamma {c}_{alpha }tau {({a}^{2}+{d}_{0}^{2})}^{2}}{4sqrt{pi }{a}^{3}{d}_{0}^{2}{r}_{0}}), respectively. The mixture proportion is determined by ({w}_{1}=sqrt{2pi {kappa }_{1}/{c}_{1}}exp Big(frac{{c}_{1}}{gamma {c}_{alpha }tau }frac{4sqrt{2}{a}^{2}{d}_{1}}{{({a}^{2}+{d}_{1}^{2})}^{1/2}}Big)) and ({w}_{0}=sqrt{2pi {kappa }_{0}/{c}_{0}}exp big(frac{{c}_{0}}{gamma {c}_{alpha }tau }frac{8sqrt{pi }{a}^{3}{d}_{0}^{2}{r}_{0}}{3({a}^{2}+{d}_{0}^{2})}big)). See Supplementary Information for the details of the mathematical derivation of Eq. (28) and Eq. (29).
Further information on research design is available in the Nature Research Reporting Summary linked to this article.
The local field potential data used in this study are available on Zenodo86 (https://doi.org/10.5281/zenodo.6806648). Source data are provided with this paper.
The code for simulations and analyses of the spiking neural circuit model87 (https://doi.org/10.5281/zenodo.6806437) and the code for fractional Hamiltonian Monte Carlo sampling88 (https://doi.org/10.5281/zenodo.6799461) are available without restrictions on Github.
Softky, W. R. & Koch, C. The highly irregular firing of cortical cells is inconsistent with temporal integration of random EPSPs. J. Neurosci. 13, 334–350 (1993).
CAS  PubMed  PubMed Central  Article  Google Scholar 
Tomko, G. J. & Crapper, D. R. Neuronal variability: non-stationary responses to identical visual stimuli. Brain Res. 79, 405–418 (1974).
CAS  PubMed  Article  Google Scholar 
Churchland, M. M. et al. Stimulus onset quenches neural variability: a widespread cortical phenomenon. Nat. Neurosci. 13, 369–378 (2010).
CAS  PubMed  PubMed Central  Article  Google Scholar 
Pouget, A., Beck, J. M., Ma, W. J. & Latham, P. E. Probabilistic brains: knowns and unknowns. Nat. Neurosci. 16, 1170–1178 (2013).
CAS  PubMed  PubMed Central  Article  Google Scholar 
Fiser, J., Berkes, P., Orbán, G. & Lengyel, M. Statistically optimal perception and learning: from behavior to neural representations. Trends Cogn. Sci. 14, 119–130 (2010).
PubMed  PubMed Central  Article  Google Scholar 
Mach, E. Contributions to the Analysis of the Sensations (ed. Williams, C. M.) (Open Court Publishing Co., 1980).
Von Helmholtz, H. Physiological Optics, Vol. III: The Perceptions of Vision (ed. Southall, J. P.) Vol. 3 (Optical Society of America, 1925).
Knill, D. C. & Richards, W. (eds.) Perception as Bayesian Inference (Cambridge University Press, 1996).
Yuille, A. & Kersten, D. Vision as Bayesian inference: analysis by synthesis? Trends Cogn. Sci. 10, 301–308 (2006).
PubMed  Article  Google Scholar 
Griffiths, T. L., Steyvers, M. & Tenenbaum, J. B. Topics in semantic representation. Psychol. Rev. 114, 211 (2007).
PubMed  Article  Google Scholar 
Vul, E., Goodman, N., Griffiths, T. L. & Tenenbaum, J. B. One and done? Optimal decisions from very few samples. Cogn. Sci. 38, 599–637 (2014).
PubMed  Article  Google Scholar 
Wolpert, D. M. Probabilistic models in human sensorimotor control. Hum Movement Sci 26, 511–524 (2007).
Article  Google Scholar 
Ma, W. J., Beck, J. M., Latham, P. E. & Pouget, A. Bayesian inference with probabilistic population codes. Nat. Neurosci. 9, 1432–1438 (2006).
CAS  PubMed  Article  Google Scholar 
Orbán, G., Berkes, P., Fiser, J. & Lengyel, M. Neural variability and sampling-based probabilistic representations in the visual cortex. Neuron 92, 530–543 (2016).
PubMed  PubMed Central  Article  CAS  Google Scholar 
Buesing, L., Bill, J., Nessler, B. & Maass, W. Neural dynamics as sampling: a model for stochastic computation in recurrent networks of spiking neurons. PLOS Comput. Biol. 7, e1002211 (2011).
ADS  MathSciNet  CAS  PubMed  PubMed Central  Article  Google Scholar 
Moreno-Bote, R., Knill, D. C. & Pouget, A. Bayesian sampling in visual perception. Proc. Natl. Acad. Sci. USA 108, 12491–12496 (2011).
ADS  CAS  PubMed  PubMed Central  Article  Google Scholar 
Hoyer, P. & Hyvärinen, A. In Advances in Neural Information Processing Systems (eds. Becker, S., Thrun, S. & Obermayer, K.) Vol. 15 (MIT Press, 2003).
Haefner, R. M., Berkes, P. & Fiser, J. Perceptual decision-making as probabilistic inference by neural sampling. Neuron 90, 649–660 (2016).
CAS  PubMed  Article  Google Scholar 
Zhang, W.-H., Lee, T. S., Doiron, B. & Wu, S. Distributed sampling-based Bayesian inference in coupled neural circuits. Preprint at bioRxiv https://doi.org/10.1101/2020.07.20.212126 (2020).
He, B. J. Scale-free brain activity: past, present, and future. Trends Cogn. Sci. 18, 480–487 (2014).
PubMed  PubMed Central  Article  Google Scholar 
Donoghue, T. et al. Parameterizing neural power spectra into periodic and aperiodic components. Nat. Neurosci. 23, 1655–1665 (2020).
CAS  PubMed  PubMed Central  Article  Google Scholar 
Townsend, R. G. & Gong, P. Detection and analysis of spatiotemporal patterns in brain activity. PLOS Comput. Biol. 14, e1006643 (2018).
ADS  PubMed  PubMed Central  Article  CAS  Google Scholar 
Muller, L., Chavane, F., Reynolds, J. & Sejnowski, T. J. Cortical travelling waves: mechanisms and computational principles. Nat. Rev. Neurosci. 19, 255 (2018).
CAS  PubMed  PubMed Central  Article  Google Scholar 
Ruderman, D. L. & Bialek, W. Statistics of natural images: scaling in the woods. Phys. Rev. Lett. 73, 814 (1994).
ADS  CAS  PubMed  Article  Google Scholar 
Munn, B. & Gong, P. Critical dynamics of natural time-varying images. Phys. Rev. Lett. 121, 058101 (2018).
ADS  CAS  PubMed  Article  Google Scholar 
MacKay, D. J. Information Theory, Inference and Learning Algorithms (Cambridge University Press, 2003).
Sanborn, A. N. & Chater, N. Bayesian brains without probabilities. Trends Cogn. Sci. 20, 883–893 (2016).
PubMed  Article  Google Scholar 
Klages, R., Radons, G. & Sokolov, I. M. Anomalous Transport: Foundations and Applications (John Wiley & Sons, 2008).
Ponce-Alvarez, A., Thiele, A., Albright, T. D., Stoner, G. R. & Deco, G. Stimulus-dependent variability and noise correlations in cortical MT neurons. Proc. Natl. Acad. Sci. USA 110, 13162–13167 (2013).
ADS  CAS  PubMed  PubMed Central  Article  Google Scholar 
Fiebelkorn, I. C. & Kastner, S. A rhythmic theory of attention. Trends Cogn. Sci. 23, 87–101 (2019).
PubMed  Article  Google Scholar 
Brascamp, J. W., Klink, P. C. & Levelt, W. J. The ‘laws’ of binocular rivalry: 50 years of Levelt’s propositions. Vision Res. 109, 20–37 (2015).
CAS  PubMed  Article  Google Scholar 
Laquitaine, S. & Gardner, J. L. A switching observer for human perceptual estimation. Neuron 97, 462–474 (2018).
CAS  PubMed  Article  Google Scholar 
Caruso, V. C. et al. Single neurons may encode simultaneous stimuli by switching between activity patterns. Nat. Commun. 9, 1–16 (2018).
ADS  CAS  Article  Google Scholar 
Rich, E. L. & Wallis, J. D. Decoding subjective decisions from orbitofrontal cortex. Nat. Neurosci. 19, 973–980 (2016).
CAS  PubMed  PubMed Central  Article  Google Scholar 
Meister, M. L. R., Hennig, J. A. & Huk, A. C. Signal multiplexing and single-neuron computations in lateral intraparietal area during decision-making. J. Neurosci. 33, 2254–2267 (2013).
CAS  PubMed  PubMed Central  Article  Google Scholar 
Horvát, S. et al. Spatial embedding and wiring cost constrain the functional layout of the cortical network of rodents and primates. PLOS Biology 14, e1002512 (2016).
PubMed  PubMed Central  Article  CAS  Google Scholar 
Levy, R. B. & Reyes, A. D. Spatial profile of excitatory and inhibitory synaptic connectivity in mouse primary auditory cortex. J. Neurosci. 32, 5609–5619 (2012).
CAS  PubMed  PubMed Central  Article  Google Scholar 
Okun, M. & Lampl, I. Instantaneous correlation of excitation and inhibition during ongoing and sensory-evoked activities. Nat. Neurosci. 11, 535–537 (2008).
CAS  PubMed  Article  Google Scholar 
Garg, A. K., Li, P., Rashid, M. S. & Callaway, E. M. Color and orientation are jointly coded and spatially organized in primate primary visual cortex. Science 364, 1275–1279 (2019).
ADS  CAS  PubMed  PubMed Central  Article  Google Scholar 
Gu, Y., Qi, Y. & Gong, P. Rich-club connectivity, diverse population coupling, and dynamical activity patterns emerging from local cortical circuits. PLOS Comput. Biol. 15, 1–34 (2019).
Article  CAS  Google Scholar 
Viswanathan, G. M. et al. Lévy flight search patterns of wandering albatrosses. Nature 381, 413–415 (1996).
ADS  CAS  Article  Google Scholar 
Viswanathan, G. et al. Optimizing the success of random searches. Nature 401, 911–914 (1999).
ADS  CAS  PubMed  Article  Google Scholar 
Harris, T. H. et al. Generalized Lévy walks and the role of chemokines in migration of effector CD8+ T cells. Nature 486, 545–548 (2012).
ADS  CAS  PubMed  PubMed Central  Article  Google Scholar 
Shlesinger, M. F., West, B. J. & Klafter, J. Lévy dynamics of enhanced diffusion: application to turbulence. Phys. Rev. Lett. 58, 1100–1103 (1987).
ADS  MathSciNet  CAS  PubMed  Article  Google Scholar 
Liu, Y., Long, X., Martin, P. R., Solomon, S. G. & Gong, P. Lévy walk dynamics explain gamma burst patterns in primate cerebral cortex. Commun. Biol. 4, 739 (2021).
PubMed  PubMed Central  Article  Google Scholar 
Pfeiffer, B. E. & Foster, D. J. Autoassociative dynamics in the generation of sequences of hippocampal place cells. Science 349, 180–183 (2015).
ADS  CAS  PubMed  Article  Google Scholar 
Andrieu, C., De Freitas, N., Doucet, A. & Jordan, M. I. An introduction to MCMC for machine learning. Mach. Learn. 50, 5–43 (2003).
MATH  Article  Google Scholar 
Zaburdaev, V., Denisov, S. & Klafter, J. Lévy walks. Rev. Modern Phys. 87, 483 (2015).
ADS  MathSciNet  CAS  Article  Google Scholar 
Berg, R. W. & Kleinfeld, D. Rhythmic whisking by rat: retraction as well as protraction of the vibrissae is under active muscular control. J. Neurophysiol. 89, 104–117 (2003).
PubMed  Article  Google Scholar 
Festa, D., Aschner, A., Davila, A., Kohn, A. & Coen-Cagli, R. Neuronal variability reflects probabilistic inference tuned to natural image statistics. Nat. Commun. 12, 1–11 (2021).
Article  CAS  Google Scholar 
Chen, G. & Gong, P. A spatiotemporal mechanism of visual attention: superdiffusive motion and theta oscillations of neural population activity patterns. Sci. Adv. 8, eabl4995 (2022).
ADS  PubMed  PubMed Central  Article  Google Scholar 
Helfrich, R. F. et al. Neural mechanisms of sustained attention are rhythmic. Neuron 99, 854–865 (2018).
CAS  PubMed  PubMed Central  Article  Google Scholar 
Xue, M., Atallah, B. V. & Scanziani, M. Equalizing excitation–inhibition ratios across visual cortical neurons. Nature 511, 596–600 (2014).
ADS  CAS  PubMed  PubMed Central  Article  Google Scholar 
Paczuski, M., Maslov, S. & Bak, P. Avalanche dynamics in evolution, growth, and depinning models. Phys. Rev. E 53, 414 (1996).
ADS  CAS  Article  Google Scholar 
Matsuoka, K. Sustained oscillations generated by mutually inhibiting neurons with adaptation. Biol. Cybernet. 52, 367–376 (1985).
MathSciNet  CAS  MATH  Article  Google Scholar 
Moldakarimov, S., Rollenhagen, J. E., Olson, C. R. & Chow, C. C. Competitive dynamics in cortical responses to visual stimuli. J. Neurophysiol. 94, 3388–3396 (2005).
PubMed  Article  Google Scholar 
Neal, R. M. In Handbook of Markov Chain Monte Carlo (2011).
Aitchison, L. & Lengyel, M. The Hamiltonian brain: efficient probabilistic inference with excitatory-inhibitory neural circuit dynamics. PLOS Comput. Biol. 12, e1005186 (2016).
ADS  PubMed  PubMed Central  Article  CAS  Google Scholar 
Savin, C. & Denève, S. In Advances in Neural Information Processing Systems (2014).
Girolami, M. & Calderhead, B. Riemann manifold Langevin and Hamiltonian Monte Carlo methods. J. R. Stat. Soc. B 73, 123–214 (2011).
MathSciNet  MATH  Article  Google Scholar 
Wozny, D. R., Beierholm, U. R. & Shams, L. Probability matching as a computational strategy used in perception. PLOS Comput. Biol. 6, 1–7 (2010).
MathSciNet  Article  CAS  Google Scholar 
Laing, C. R. & Chow, C. C. A spiking neuron model for binocular rivalry. J. Comput. Neurosci. 12, 39–53 (2002).
PubMed  Article  Google Scholar 
Brascamp, J. W., van Ee, R., Pestman, W. R. & van den Berg, A. V. Distributions of alternation rates in various forms of bistable perception. J. Vis. 5, 1–1 (2005).
Article  Google Scholar 
Moreno-Bote, R., Rinzel, J. & Rubin, N. Noise-induced alternations in an attractor network model of perceptual bistability. J. Neurophysiol. 98, 1125–1139 (2007).
PubMed  Article  Google Scholar 
Echeveste, R., Aitchison, L., Hennequin, G. & Lengyel, M. Cortical-like dynamics in recurrent circuits optimized for sampling-based probabilistic inference. Nat. Neurosci. 23, 1138–1149 (2020).
CAS  PubMed  PubMed Central  Article  Google Scholar 
Otero-Millan, J., Troncoso, X. G., Macknik, S. L., Serrano-Pedraza, I. & Martinez-Conde, S. Saccades and microsaccades during visual fixation, exploration, and search: foundations for a common saccadic generator. J. Vis. 8, 21–21 (2008).
PubMed  Article  Google Scholar 
Brockmann, D. & Geisel, T. The ecology of gaze shifts. Neurocomputing 32, 643–650 (2000).
Article  Google Scholar 
Lisman, J. & Jensen, O. The theta-gamma neural code. Neuron 77, 1002–1016 (2013).
CAS  PubMed  PubMed Central  Article  Google Scholar 
Gelfand, A. E. & Smith, A. F. Sampling-based approaches to calculating marginal densities. J. Amer. Statist. Assoc. 85, 398–409 (1990).
MathSciNet  MATH  Article  Google Scholar 
Bak, P., Tang, C. & Wiesenfeld, K. Self-organized criticality. Phys. Rev. A 38, 364 (1988).
ADS  MathSciNet  CAS  MATH  Article  Google Scholar 
di Santo, S., Villegas, P., Burioni, R. & Muñoz, M. A. Landau–Ginzburg theory of cortex dynamics: scale-free avalanches emerge at the edge of synchronization. Proc. Natl. Acad. Sci. USA 115, E1356–E1365 (2018).
PubMed  PubMed Central  Article  CAS  Google Scholar 
Fontenele, A. J. et al. Criticality between cortical states. Phys. Rev. Lett. 122, 208101 (2019).
ADS  CAS  PubMed  Article  Google Scholar 
Chen, G. & Gong, P. Computing by modulating spontaneous cortical activity patterns as a mechanism of active visual processing. Nat. Commun. 10, 4915 (2019).
ADS  PubMed  PubMed Central  Article  CAS  Google Scholar 
Ackley, D. H., Hinton, G. E. & Sejnowski, T. J. A learning algorithm for Boltzmann machines. Cogn. Sci. 9, 147–169 (1985).
Article  Google Scholar 
Hinton, G. E., Osindero, S. & Teh, Y.-W. A fast learning algorithm for deep belief nets. Neural Comput. 18, 1527–1554 (2006).
MathSciNet  PubMed  MATH  Article  Google Scholar 
Neal, R. M. Sampling from multimodal distributions using tempered transitions. Stat. Comput. 6, 353–366 (1996).
Article  Google Scholar 
Georgopoulos, A., Kalaska, J., Caminiti, R. & Massey, J. On the relations between the direction of two-dimensional arm movements and cell discharge in primate motor cortex. J. Neurosci. 2, 1527–1537 (1982).
CAS  PubMed  PubMed Central  Article  Google Scholar 
Pouget, A., Zhang, K., Deneve, S. & Latham, P. E. Statistically efficient estimation using population coding. Neural Comput. 10, 373–401 (1998).
CAS  PubMed  Article  Google Scholar 
Nolan, J. P. Univariate Stable Distributions: Models for Heavy-Tailed Data (Springer International Publishing, 2020).
Ye, N. & Zhu, Z. Stochastic fractional Hamiltonian Monte Carlo. In International Joint Conference on Artificial Intelligence (2018).
West, B. J. Colloquium: Fractional calculus view of complexity: a tutorial. Rev. Mod. Phys. 86, 1169 (2014).
ADS  Article  Google Scholar 
Wardak, A. & Gong, P. Fractional diffusion theory of balanced heterogeneous neural networks. Phys. Rev. Res. 3, 013083 (2021).
CAS  Article  Google Scholar 
Şimşekli, U. Fractional Langevin Monte Carlo: exploring Lévy driven stochastic differential equations for Markov chain Monte Carlo. Proc. Mach. Learn. Res. 70, 3200–3209 (2017).
Google Scholar 
Li, C. & Zeng, F. Finite difference methods for fractional differential equations. Int. J. Bifurcation Chaos 22, 1230014 (2012).
ADS  MathSciNet  MATH  Article  Google Scholar 
Fung, C. C. A., Wong, K. Y. M. & Wu, S. A moving bump in a continuous manifold: a comprehensive study of the tracking dynamics of continuous attractor neural networks. Neural Comput. 22, 752–792 (2010).
MathSciNet  PubMed  MATH  Article  Google Scholar 
Qi, Y. & Gong, P. Fractional neural sampling as a theory of spatiotemporal probabilistic computation in neural circuits, local field potentials. Zenodo, https://doi.org/10.5281/zenodo.6806648 (2022).
Qi, Y. & Gong, P. Fractional neural sampling as a theory of spatiotemporal probabilistic computation in neural circuits, fractional Hamiltonian Monte Carlo sampler. Github, https://doi.org/10.5281/zenodo.6806437 (2022).
Qi, Y. & Gong, P. Fractional neural sampling as a theory of spatiotemporal probabilistic computation in neural circuits, SpikeNet. Github, https://doi.org/10.5281/zenodo.6799461 (2022).
Download references
This work was supported by Australian Research Council Grants DP160104316 (P.G.).
School of Physics, University of Sydney, Sydney, NSW, 2006, Australia
Yang Qi & Pulin Gong
Institute of Science and Technology for Brain-Inspired Intelligence, Fudan University, Shanghai, 200433, China
Yang Qi
Key Laboratory of Computational Neuroscience and Brain-Inspired Intelligence (Fudan University), Ministry of Education, Shanghai, China
Yang Qi
MOE Frontiers Center for Brain Science, Fudan University, Shanghai, 200433, China
Yang Qi
You can also search for this author in PubMed Google Scholar
You can also search for this author in PubMed Google Scholar
Conceptualization: Y.Q. and P.G.; methodology: Y.Q. and P.G.; investigation: Y.Q. and P.G.; writing—original draft: Y.Q. and P.G.; writing—review & editing: Y.Q. and P.G.; funding acquisition: P.G.; supervision: P.G.
Correspondence to Pulin Gong.
The authors declare no competing interests.
Nature Communications thanks Biyu He and the other anonymous reviewer(s) for their contribution to the peer review of this work
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
Reprints and Permissions
Qi, Y., Gong, P. Fractional neural sampling as a theory of spatiotemporal probabilistic computations in neural circuits. Nat Commun 13, 4572 (2022). https://doi.org/10.1038/s41467-022-32279-z
Download citation
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41467-022-32279-z
Anyone you share the following link with will be able to read this content:
Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.
Focus
Advertisement
Advanced search
Nature Communications (Nat Commun) ISSN 2041-1723 (online)
© 2022 Springer Nature Limited
Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

source

WHAT DOES FAMILY MEAN? EIGHT LOCAL FILMMAKERS OFFER UNIQUE PERSPECTIVE DURING FREE SCREENING EVENT AT MDC'S TOWER THEATER MIAMI AUGUST 25 – Miami Dade College

New “Family Commissions” program invested a total of $120,000 into local filmmakers to develop original new works
MIAMI (August 9, 2022) – Using vastly different storylines, eight Miami filmmakers will showcase their own perspective on what “family” means during a free screening at Miami Dade College’s (MDC) Tower Theater Miami from 7-9 p.m. on Thursday, August 25. The films are part of “Family Commissions,” a program of eight short films presented by Oolite Arts and MDC’s Miami Film Festival. Co-funded by the Lynn and Louis Wolfson II Family Foundation, each filmmaker received a share of $120,000 to produce original works using the theme of family.
“Alongside our friends at Oolite Arts, we are thrilled to be supporting the work of these immensely talented local filmmakers and helping them take this next step in advancing their promising careers,” said Nicolas Calzada, interim executive director of MDC’s Miami Film Festival and Tower Theater Miami. “These films are both entertaining and impressive. We can’t encourage the community enough to come out and experience the unique stories that these emerging filmmakers have to share.”
The stories include two teenagers’ quest to find an ancient good luck charm in hopes of saving their family booths and a flea market from closure; a first-person account of a filmmaker’s quarter-life crisis brought on by his ever-changing family dynamics, and the escapades of a young man who will stop at nothing to recover a stolen family heirloom.
“Our talented community of indie filmmakers has done it again, turning a topic we all embrace  – family – into an on-screen investigation of the human condition,” said Dennis Scholl, Oolite Arts’ president & CEO. “We’re delighted to support our local filmmakers by bringing their stories to the big screen in partnership with the Miami Film Festival and the Lynn and Louis Wolfson II Family Foundation.”
At the August 25 screening event, attendees will also have an opportunity to meet the eight filmmakers who had up to six months to work on their respective films. They are:
To RSVP to the FREE screening on August 25, visit https://www.towertheatermiami.com/event/mff-oolite-arts-family-commissions.
CALENDAR NOTICE
“Family Commissions” Works-in-Progress Screening
Eight short films by local filmmakers centered on the theme family
Presented by Oolite Arts and MDC’s Miami Film Festival
7-9 p.m. Thursday, August 25 at Tower Theater Miami
1508 SW 8th Street, Miami, FL 33135
RSVP: https://www.towertheatermiami.com/event/mff-oolite-arts-family-commissions
About Oolite Arts
Oolite Arts helps Miami-based artists advance their careers and inspires the cultural community to engage with their work. Established in 1984, the nonprofit is both a community and a resource, providing visual artists with the studio space, exhibition opportunities and financial support they need to experiment, grow and enrich the city. Through its educational programming, Oolite Arts helps Miamians learn about contemporary art and develop their own artistic skills. Coming in 2024—in time for the organization’s 40th anniversary—Oolite Arts will open a new campus in the City of Miami that will help build Miami’s next creative chapter. For more, visit oolitearts.org.
Exhibitions and programs at Oolite Arts are made possible with the support of the Miami-Dade County Department of Cultural Affairs and the Cultural Affairs Council; the Miami-Dade County Mayor and Board of County Commissioners; the Miami Beach Mayor and City Commissioners; the State of Florida, Department of State, Division of Arts and Culture and the Florida Council on Arts and Culture; the National Endowment for the Arts; the Jorge M. Perez Family Foundation at the Miami Foundation, the Lynn and Louis Wolfson II Family Foundation, and the John S. and James L. Knight Foundation. For more information, visit oolitearts.org. Follow @oolitearts on social media.
About Miami Dade College’s Miami Film Festival   
Miami Film Festival is an international film event that serves as a major film showcase for world cinema. Considered the preeminent film festival for highlighting Ibero-American cinema in the U.S, Miami Film Festival has become renowned for championing films made by the South Florida community, first-time feature filmmakers, and International Feature Film submissions to the Academy Awards. The annual Festival, produced and presented by Miami Dade College, attracts more than 60,000 audience members and more than 400 filmmakers, producers, talent, and industry professionals. Since 1984, the Festival has screened over 2,500 films from more than 75 countries and given out more than $2 million in cash awards to filmmakers and distributors. Miami Film Festival is the oldest major festival housed in a university or college and also operates a year-round art cinema, Tower Theater Miami and Miami Film Festival GEMS in November. For more information, visit miamifilmfestival.com or call 305-237-FILM (3456).
Oolite Arts + Miami Film Festival Media Relations Contacts:
Andrea Salazar, Pinzur Communications | 954-756-0652 | [email protected]
Rachel Pinzur, Pinzur Communications | 305-725-2875 | [email protected]

source

The 10 Best Restaurants In Gainesville, FL – Tasting Table

It’s great to be a Florida Gator. There’s the school spirit, warm weather, and the prestigious university accolades (one of the top public universities in the country). The University of Florida (UF) is in the small city of Gainesville, Florida, and the local food scene is thriving. Charming brunch spots perfect for your Instagram pictures, casual dinner places ideal for students balling on a budget, and cuisines from many different cultures are around this town. Sadly, some Gainesville classics like The Swamp Restaurant, Leonardo’s By The Slice, Larry’s Giant Subs, and Felipe’s Mexican Taqueria have closed due to the pandemic, reports The Gainesville Sun (though The Swamp is luckily being rebuilt in a different part of town). These places and others will all be missed. If you graduated from UF a few years ago and are wondering what the best spots in town are (that are still open or new spots) and want to avoid rolling up to Felipe’s only to find it closed, this article is for you.
If you’re going to be a new student at UF this fall and are looking to try some spots not in the Reitz Union or in Midtown, here are the best local food spots you need to check out with your parents during a move-in weekend or with friends. Exploring the city’s best places will make you feel more at home, and you can try some amazing food. Here are the 10 best restaurants in Gainesville, Florida.

If you walk through Downtown Gainesville at 11 a.m. and notice people with mimosas and plates full of delectable brunch tacos, eggs, and avocado toast, you’re looking at Original American Kitchen (OAK). This place has delicious brunch and dinner with elevated classics like wild mushroom ravioli, the best grilled cheese and tomato soup ever (no, seriously), fried green tomato sliders, and bottomless mimosas. Satisfy your sweet tooth with desserts like homemade bread pudding, pecan pie, and fried Oreos. You can sit either inside, outside, or at the bar if you choose.
Wherever you sit, you can bet the decor will be rustic and sweet. Tons of unique mirrors create a gallery wall at the back of the restaurant, a wall on the right features mason jars filled with flowers, and a large chalkboard wall has doodles and menu details. OAK is an absolute delight that will leave you coming back for more delicious meals and photos with friends. It’s also conveniently located Downtown, so once those mimosas and brunch food have filled you up, take a stroll around town and appreciate historic Gainesville. Pop in a record store, get an iced coffee from Wyatt’s Coffee, and scope out where you want to go for dinner later (or where to order from). What a perfect Saturday morning.

This restaurant is for all the folks who look forward to Publix’s $5 Wednesday sushi rolls (yum). One of the fancier dinner spots in town, this place has some of the highest quality sushi you’ll ever have (it also features a vegetarian menu, which provides options for a wide range of people). Dragonfly Sushi & Sake Company (commonly called Dragonfly Sushi) is not budget-friendly like most other places on this list, but it’s perfect for your special occasions. For example, when your family pops into town for parents’ weekend, graduation, or a football game, go to Dragonfly Sushi with them. It’ll be nice to get an elevated meal after months of Panda Express, Chipotle, and Wawa (absolutely no shade to any of those places, though). If you go with family, taking them to a local spot will make you feel like a local and probably impress them.
Set in Downtown, Dragonfly Sushi sits on an ideal corner right in the middle of everything. After getting absolutely stuffed from sushi, seafood dishes, and a few drinks, pop next door to The Hyppo Gourmet Ice Pops for some light and refreshing fruit pops. They come in unique flavors like blueberry lavender lemonade and mango cheesecake. Or, if you’re more of a dinner-and-a-show type, check out what’s playing at the Hippodrome Theatre one block away from the restaurant, as there is always a new show to see there.

If you’re looking for authentic Mexican food, La Tienda has the best in Gainesville. If you’re a new student looking to impress the friends you just made by taking them to a great local Mexican restaurant that’s not Chuy’s (sorry Chuy’s, we really do love you), go to La Tienda. You won’t have to stress about navigating the mildly stressful drive-through situation at El Indio (it has delicious Mexican food, but with only a few tables outside, it’s not a sit-down place). The food at La Tienda is affordable and offers high value. Though usually packed with hungry people ready for food, it is worth scoping out a table here because the tacos are unmatched.
This spot is on SW 13th Street past Shands Hospital, so it’s not super walkable from campus, but it’s a short two-minute drive (or hop on a scooter to get there quickly). We recommend getting either the enchiladas, soft tacos, or a burrito. La Tienda also has a vegetarian menu, so bring your veggie friends! Get an order of chips, salsa, and queso to share. You won’t regret it. La Tienda is open from 11 a.m. to 9:30 p.m. Monday to Thursday, 11 a.m. to 10 p.m. Friday and Saturday, and 11 a.m. to 9 p.m. on Sunday. If you don’t have time to stop here, you can always order the food to-go or opt for delivery!

Satchel’s Pizza is one of those places that you almost don’t believe exists until you see it for yourself. Mosaic tiling covers every inch of this pizza joint, from the floor to the ceiling. Tiles line the inside spaces, the exterior, the patio area, all the outdoor tables, and even the converted van. Getting a van seat is a coveted experience and a great photo opportunity. However, Satchel’s isn’t just about the quirky aesthetics — the food is incredible. If you’re a New York or New Jersey native who has found themselves in Florida, Satchel’s Pizza might have the best pizza in the state. Not only is the pie high-quality, but the sodas are homemade, and the desserts are decadent. Even if you’re not hungry for dessert, you have to ask for the dessert menu. It actually comes in a View-Master toy that you hold up to your eyes and use to click through images of the dessert options.
This restaurant is almost on the edge of town, near the airport, but it is worth the extra ten minutes of driving. Satchel’s also frequently hosts live music, so if you want the best pizza in town for an affordable price, a unique and fun atmosphere that will heal your inner child, and an experience you’ll be talking about for weeks to come, head to Satchel’s Pizza.

Just because Gainesville is two hours from the coast doesn’t mean you can’t find great seafood here. Harry’s Seafood Bar and Grille is a New Orleans-inspired spot with delicious seafood and steaks perfect for a celebratory dinner with family or friends. This restaurant is one of the more upscale ones in town, so it’s ideal for a birthday dinner, graduation celebration, or when family comes into town to visit. Located Downtown (can you tell how many good spots are Downtown?), Harry’s has signature crab cakes that are authentic and to-die-for. The New Orleans ribeye is tender and juicy, and the French-baked scallops are delicious.
If you don’t mind eating dinner a little earlier than the average person or are a big “linner” enthusiast, take advantage of Harry’s happy hour from 2 p.m. to 7 p.m. All draft beer is $3, premium cocktails and glasses of wine are $4, and many more discounts await. From the bar menu, we recommend the Bayou Bloody Mary for those who love a little kick or The Front Porch (made with Southern iced tea, bourbon, orange liqueur, and muddled mint) for people craving a sweeter beverage (it’s a spiked sweet tea). For dessert, sink into the mouthwatering bananas foster, key lime pie, or Oreo beignets. No matter what you order from the menu, it will be delicious at Harry’s Seafood Bar and Grille.

Don’t be fooled by the name. Flaco’s Cuban Bakery is serving up top-notch lunches and dinners. The Cuban sandwiches are the perfect simple lunch, and the arepas are a surefire hangover cure. Flaco’s, as it’s lovingly referred to, is the ideal drunk food if you’re out Downtown. After a Downtown bar crawl (Midtown is a different story), going from UC (University Club) to Arcade Bar to Cry Baby’s to Downtown Fats, end your night on a high note at Flaco’s. With a closing time of 2:30 a.m., this establishment knows that people need some delicious arepas, Cuban sandwiches, tacos, burritos, burrito bowls, or empanadas to settle their stomachs and end the night right. If you’re choosing Taco Bell instead of Flaco’s at 2 a.m., sorry, we’re judging you.
If you’re not a night owl, don’t worry because Flaco’s opens at 11 a.m., the perfect time for lunch on a sunny Gainesville day. Another great thing to note about Flaco’s is that it’s an ideal spot for students, with menu items averaging around $10. Not bad at all! If you’re out studying at a coffee shop Downtown and start to want something a bit more substantial than coffee, go to Flaco’s. You won’t be disappointed.

If the line wrapped around Luke’s New York Bagel Shop every morning is any indication, this small Gainesville bagel shop serves the most authentic New York bagels and breakfast sandwiches around. The shop exterior recently got a fire engine red facelift, so if you’re driving around trying to find the shop, you can’t miss it. Be sure to park in the neighboring lot, as the parking lot at Luke’s will feature a line of people waiting for their coffee and bagels. If you want the authentic Gainesville experience, park your car for free in the Depot Park parking lot right across the street. Then, walk over to Luke’s, grab your food, and set up a blanket in idyllic Depot Park to sit and make a picnic out of breakfast (don’t worry if you forgot a blanket; there are plenty of picnic tables in the park).
If you want Luke’s full range of bagel flavors, cream cheeses, and breakfast sandwiches, make a point to get there early, before 9 a.m. The breakfast spot opens at 7 a.m. and closes at 1 p.m., and if you’re an early riser anyway, the line isn’t too long at opening time. Around 8 a.m. is when it is the longest. But no matter the length of the line, the food will be worth it. Authentic, huge New York bagels with generous amounts of cream cheese await.

Mi Apá Latin Café (casually referred to as Mi Apá) is a sensation. Located right off NW 34th Street, Mi Apá is perfect for an after-class pick-me-up. If you’re going there for breakfast, get the cafe con leche and stuffed arepa (with scrambled eggs, ham, and cheese inside). If you’re popping by for lunch or dinner, we recommend ordering one of the entrees and definitely getting rice and beans on the side (if they don’t come with your meal already). The menu is vast but not overwhelming, meaning everyone will find something they like. Mi Apá’s prices are very student-friendly, and the drive-through line is efficient and easy, so if you’re a busy student on the go, this restaurant is perfect.
If you’re looking for a place to rejuvenate after some shopping on 34th Street, Mi Apá is right near all the shops and is the perfect place to stop and get a bite to eat. If you want to do something fun with your friends after lunch at Mi Apá, stop by Corks & Colors Studio to paint some pottery (located right in the same plaza). All in all, Mi Apá is a great local spot to check out if you’re on 34th Street or want to find good Cuban food in Gainesville.

Weekend brunch in Gainesville can mean you’re fighting everyone else to get a reservation somewhere, but the short wait at Afternoon is always worth it. This brunch spot is hidden away on NW 10th Avenue, but the quality is well-known to most people in town. Afternoon is a Pinterest girl’s dream. Mint green accents in the decor, rose gold fixtures, a record player churning out tunes, and a beautiful feature wall are just a few things that make the vibe in this restaurant excellent. However, the good vibes don’t stop at the decor; the food here is also phenomenal. The coffee options are top-tier, and the unique brunch options are incredible. For drinks apart from coffee, we recommend the matcha tea latte, a spritz (an alcoholic fizzy cocktail), or a butterfly flower lemonade. For your main course, some tried-and-true items are the Dutch baby, mushroom tartine, poached eggs, shakshuka, or the classic breakfast.
After your brunch at Afternoon, keep your car parked and walk over to the other side of the plaza into Serpentine Plants + Provisions. Owned by the same folks as Afternoon, this plant and gift shop is just as charming as the restaurant, and you’ll find some great coffee table books, wine, earrings, candle holders, and plants here.

Football isn’t for everybody, and that’s okay. If you’re not a football fan, you might not know the name Steve Spurrier. Spurrier is one of the most famous University of Florida football players in history, and he even won the legendary Heisman Trophy in 1966, per Heisman Trophy. After his playing days, he went on to be a college football coach. He eventually came back and coached at his alma mater and guided the Gators to win the SEC championship in 1990, 1991, 1993, 1994, 1995, 1996, and 2000. Spurrier is basically a Gainesville legend.
Steve Spurrier has since retired from coaching, but he wasn’t done with Gainesville just yet. In 2021, he opened Spurrier’s Gridiron Grille, a classy restaurant in the Celebration Point plaza. He does food as well as football. This upscale place offers to-die-for cocktails and appetizers (the orange and blue shrimp are great), and the entrees are just as unforgettable. This eatery is a bit pricey for students, but it’s the perfect place to bring your family members who were there to witness Steve Spurrier’s career. The food is excellent, and we recommend the Emory & Henry Seafood Pasta, Spurrier’s bacon cheeseburger, or a steak cooked to your liking. Desserts include cheesecake, lemon tarts, and chocolate chip cookies. The bar menu is extensive, even featuring signature cocktails and mocktails. We recommend the raspberry lemon drop martini or the spicy mango mocktail. Spurrier’s is also an ideal spot for those who are gluten-free.

source

Cutting-edge American technology given away to Chinese company : Consider This from NPR – NPR

MARY LOUISE KELLY, HOST:
Just north of Seattle near the shores of Puget Sound sits an empty warehouse in Mukilteo, Wash.
CHRIS HOWARD: We used to have 10 shipping containers here. There were empty containers back here, customers and clients coming for visits.
KELLY: Chris Howard used to work in that warehouse with more than a dozen other engineers and researchers for an American company called UniEnergy. The name is still on the sign out front. They were building a battery – not just any battery, something called a vanadium redox flow battery. It was about the size of a refrigerator. And Howard and the rest of the employees thought it was going to change the world.
HOWARD: It was more than a job. It was a lot of blood, sweat and tears into developing a product that we were really excited about and really proud of.
KELLY: Unlike batteries in cell phones, even cars, these batteries could charge and discharge energy for as long as 30 years. And this particular design seemed to hold enough energy to power a house. Researchers pictured people plunking them down next to their air conditioners, attaching solar panels to them, and everyone living happily ever after off the grid.
HOWARD: It was beyond promise. We were seeing it functioning as designed, as expected.
KELLY: They thought the batteries would be the next great American success story. That is not what happened. Today, that warehouse sits, shuttered, empty. All the employees who worked there were laid off. And those batteries, they are now being made at a factory in northeastern China. CONSIDER THIS – cutting edge technology developed in an American national lab, paid for with American tax dollars, was given away to a Chinese company. Using internal government documents, an NPR investigation tells the story for the very first time and shows how the U.S. may have lost the next big thing to China yet again.
(SOUNDBITE OF MUSIC)
KELLY: From NPR, I’m Mary Louise Kelly. It’s Tuesday, August 9.
It’s CONSIDER THIS from NPR. It’s a favorite promise of politicians – keep manufacturing jobs and technology in America.
(SOUNDBITE OF ARCHIVED RECORDING)
BARACK OBAMA: Thank you. Please, everybody, have a seat.
KELLY: Let’s go back to Feb. 22, 2011. It was a Tuesday, and then-President Barack Obama was in Cleveland, Ohio.
(SOUNDBITE OF ARCHIVED RECORDING)
OBAMA: The world wants American goods, and my administration is going to go to bat for America’s businesses around the world. You should know that.
(APPLAUSE)
KELLY: After shouting out LCD screens and fiber optic cables, Obama turned to an up and coming technology that seemed to show promise.
(SOUNDBITE OF ARCHIVED RECORDING)
OBAMA: And I have no idea what this is – vanadium redox fuel cells.
(LAUGHTER)
OBAMA: That’s one of the coolest things I’ve ever said out loud.
(LAUGHTER)
KELLY: Researchers at the federal government’s Pacific Northwest National Lab surely would have agreed. For the past five years, they had been developing vanadium redox technology, getting it ready for commercial use. And this came at a time when the Obama administration was making a major push to reinvigorate American manufacturing.
(SOUNDBITE OF ARCHIVED RECORDING)
OBAMA: So we have a huge opportunity at this moment to bring manufacturing back, but we have to seize it.
KELLY: That’s Obama at his State of the Union address the following year, 2012.
(SOUNDBITE OF ARCHIVED RECORDING)
OBAMA: Tonight, my message to business leaders is simple. Ask yourselves what you can do to bring jobs back to your country, and your country will do everything we can to help you succeed.
(APPLAUSE)
KELLY: That same year, UniEnergy Technologies was founded with a license to make the vanadium redox flow batteries developed at the National Lab. And as the company grew, claims about supporting American technology and manufacturing kept coming from the White House, regardless of which president lived there.
(SOUNDBITE OF ARCHIVED RECORDING)
DONALD TRUMP: We’re bringing manufacturing back to the United States big league. We’re reducing taxes.
KELLY: It was a particular fixation for Donald Trump during his time in office.
(SOUNDBITE OF ARCHIVED RECORDING)
TRUMP: I remember when I was growing up, I’d see made in America all over the place – a little bit made in the USA – but made in America. Everything had made in America. And we’re starting that again. And it’s happening again, as you know. We’re here today…
KELLY: And Trump’s successor would pick up that baton, too.
(SOUNDBITE OF ARCHIVED RECORDING)
PRESIDENT JOE BIDEN: And if we think we’re not in a race, well, guess what? Take a look at China. Take a look at China.
KELLY: President Biden last summer at a Mack Trucks facility in Pennsylvania.
(SOUNDBITE OF ARCHIVED RECORDING)
BIDEN: It’s a straightforward solution – support and grow more American-based companies.
(SOUNDBITE OF MUSIC)
KELLY: And yet, it was the federal government itself that let those cutting edge batteries slip away. NPR’s Laura Sullivan and Courtney Flatt from Public Radio’s Northwest News Network will pick up the story from here with their investigation into how vanadium redox flow batteries ended up being made in a factory in China.
LAURA SULLIVAN, BYLINE: The Chinese company didn’t steal this technology. It was given to them by the U.S. Department of Energy. An NPR investigation found the department allowed the technology and jobs to move overseas, violating its own licensing rules while failing to intervene on behalf of U.S. workers in multiple instances according to internal department emails. Now, China is forging ahead, investing millions into this cutting edge green technology that was supposed to help keep the U.S. and its economy out front.
JOANNE SKIEVASKI: It just is mind-boggling.
SULLIVAN: Joanne Skievaski is the vice president of finance for a U.S. company called Forever Energy that has been trying to get a license from the department to make the batteries here for more than a year.
SKIEVASKI: This is technology made from taxpayer dollars. It was invented by a national lab. And it’s deployed in China, and it’s held in China. To say it’s frustrating is an understatement.
SULLIVAN: Department of Energy officials declined NPR’s request for an interview and wouldn’t explain how technology that cost U.S. taxpayers $15 million ended up in China. But after NPR sent department officials detailed questions laying out the timeline of events, officials terminated the license it gave to the Chinese company. In a statement, officials said the department, quote, “takes American manufacturing obligations extremely seriously” and is now, quote, “undertaking an internal review of the licensing of vanadium battery technology.” The story of how this happened begins where the battery was born, three hours southwest of Seattle, in the basement of a government lab called the Pacific Northwest National Laboratory, where Courtney went to visit.
COURTNEY FLATT, BYLINE: Safety goggles on.
VINCE SPRENKLE: Yep.
FLATT: Vince Sprenkle works with energy storage here at the lab.
SPRENKLE: We’re going to go down into the redox flow battery lab.
FLATT: It was down here in 2006 that more than two dozen scientists began to suspect that a special mix of acid and electrolyte could hold unusual amounts of energy without degrading. They turned out to be right.
Do you feel kind of, like, on the cutting edge of learning about these batteries?
SPRENKLE: Yeah, we are. I mean, I think we’ve got one of the leading research groups in the country and probably the world in this technology.
FLATT: It’s because of this leading edge that when a success happens, the lab encourages scientists to go out and see if they can make and sell the inventions in the real world. The lab and the U.S. government still hold the patents because American taxpayers paid for the research. But the Department of Energy licenses the patents to scientists and companies willing to take a shot.
In this case, it took six years and millions of taxpayer dollars to discover the perfect battery recipe. Gary Yang was the lead scientist, and he was excited to see if he could make them. In 2012, he left the lab with the license in hand and started UniEnergy Technologies at the warehouse in Mukilteo, Washington.
GARY YANG: I left the lab – total legal process – and start UniEnergy Technology (ph) in Washington state.
FLATT: He hired engineers and researchers, but then he ran into trouble. He says he couldn’t find any U.S. investors.
YANG: I talked to almost all major investment bank. None of them invest in battery.
FLATT: So he turned to a Chinese businessman and a company called Dalian Rongke Power and its parent company, which agreed to invest and even help manufacture the batteries. And so began a slow shift. First, Chris Howard said, it was just some parts. Ultimately, it was the whole process.
HOWARD: Manufacturing was subsequently shifted to our sister company in China, and they would take on that role.
FLATT: In 2017, Yang and UniEnergy formalized the situation and gave Rongke Power an official sublicense, allowing the company to make the batteries.
SULLIVAN: So here’s the thing – companies can choose to manufacture in China, but in this case, Yang’s original license clearly says on page six, he has to sell batteries in the United States, and those batteries have to be, quote, “substantially manufactured” here. Yang acknowledges he didn’t do that. He was mostly selling batteries in China, and the batteries he did sell here were largely made in China. But he says in all those years, the department never raised any concerns or intervened.
Then, in 2019, Chris Howard said he and the other engineers were called to a conference room. Supervisors told them they, too, would have to go to China to work there for four months at a time at Rongke Power.
HOWARD: And that was set to be increased on the premise that there were certain government programs, Chinese government programs, that would support funding efforts. So it was unclear, certainly to myself and some of the other engineers, what the plan was.
SULLIVAN: In a statement, the department said that license monitoring is a priority and that a review of this case is underway – all of which brings us to Yang. Yang was born in China, but he is a U.S. citizen and got his Ph.D. at the University of Connecticut. He says he wanted to manufacture here, but at the time, China was doing more to encourage battery production. And he told us China could do it better.
YANG: In this field – manufacturing, engineering – China ahead of U.S.
SULLIVAN: China is ahead of the U.S.
YANG: They’re ahead of U.S. – many wouldn’t believe, engineering wise, ahead.
SULLIVAN: He says, far from helping China, the Chinese engineers were helping his U.S. employees. But you can see in several news reports at the time, it was helping China. The Chinese government launched several large demonstration projects and announced millions in funding.
FLATT: As things began to take off in China, here in the States, Yang was once again in financial trouble. So he made a decision that would again keep the technology from staying in the U.S. He transferred the license from UniEnergy to a company called Vanadis.
ROELOF PLATENKAMP: Vanadis is based in the Netherlands, and we will set up the holding company in Switzerland.
FLATT: Roelof Platenkamp is Vanadis’s founding partner. Platenkamp told NPR the company’s plan was to continue making the batteries in China and then set up a factory in Germany and eventually maybe the U.S. He says he has to manufacture in Europe because the European Union has strict rules about these things.
PLATENKAMP: I have to be a European company – yeah – or certainly a non-Chinese company in Europe.
FLATT: But the United States has these rules, too. And any transfer of a U.S. government license needs U.S. government approval.
SULLIVAN: Which Yang apparently had no trouble getting. We looked at department emails and found that last summer, on July 7, one of the top officials at UniEnergy wrote to a government manager at the Department of Energy Lab in Washington, saying they were going to make a deal with Vanadis. We believe they have the right blend of technical expertise, the official wrote. The manager wrote back that he would need confirmation. A second employee sent confirmation an hour and a half later, and the license was transferred.
Now, if anyone from the lab or the Department of Energy during that hour and a half thought to check whether Vanadis was an American company or whether it intended to manufacture in the United States is unclear. Even Vanadis’s website says it plans to make the batteries in China. Department of Energy officials told us they take all license transfer seriously and have recently closed significant loopholes. But they acknowledge their efforts rely to some extent on, quote, “good faith disclosures” by the companies, which means if companies like UniEnergy don’t say anything, the U.S. government may never know. It’s a problem government investigators found has been going on for years. In 2018, the Government Accountability Office found the department lacked resources to properly monitor its licenses, was relying on antiquated computer systems and didn’t have consistent policies across its labs.
FLATT: It was an American company, Forever Energy, that actually read the vanadium battery license and raised a red flag more than a year ago. Joanne Skievaski and others there say they repeatedly warned department officials that the UniEnergy license was not in compliance. Officials repeatedly told them it was.
SKIEVASKI: How is it that the national lab did not require U.S. manufacturing? Not only is it a violation of the license, it’s a violation to our country.
FLATT: Skievaski hopes that now that the department has revoked the license, Forever Energy will get a chance. They’re hoping to open a factory in Louisiana.
SKIEVASKI: We are ready to go with this technology.
SULLIVAN: Skievaski told us it will be hard at this point for any American company to catch up. Industry trade reports list Dalian Rongke Power as the No. 1 manufacturer of vanadium flow batteries worldwide. And the bigger question looming over all of this is whether China will stop making the batteries once an American company is granted the right to start making them.
FLATT: That may be unlikely. Chinese news reports announced this summer that China is about to bring online one of the largest battery farms in the world, hoping to set new records for energy output. The reports say the entire battery farm is built out of vanadium redux flow batteries.
(SOUNDBITE OF MUSIC)
KELLY: That was Courtney Flatt from public radio’s Northwest News Network along with NPR’s Laura Sullivan. You can read their entire investigation at npr.org.
(SOUNDBITE OF MUSIC)
KELLY: It’s CONSIDER THIS from NPR. I’m Mary Louise Kelly.
Copyright © 2022 NPR. All rights reserved. Visit our website terms of use and permissions pages at www.npr.org for further information.
NPR transcripts are created on a rush deadline by an NPR contractor. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of NPR’s programming is the audio record.
Sponsor Message
Become an NPR sponsor

source

‹ Previous Next ›