Blog

Multifactor assessments to determine the overall performance of supercritical fluid extraction from Gynura procumbens essential oil | 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: 14293 (2022)
Metrics details
Gynura procumbens is a medicinal herb that contains bioactive compounds that can relieve coughs and prevent liver cancer. Supercritical fluid extraction (SFE) was suggested as one of the techniques that can be used to extract the valuable compounds from the G. procumbens. SFE was widely applied in extracting medicinal ingredients from herbs. However, most of them were performed only at the laboratory scale. Moreover, study to increase the yield performance, economic studies and safety assessments of the SFE process were also performed; however, these tests were conducted individually. Moreover, to date, there is no integration study between all the factors stated for determining the overall performance of SFE with herbs specifically G. procumbens. The integration between all the factors is beneficial because the data on the overall performance can assist in developing the SFE process with G. procumbens at the pilot or industrial scale. Therefore, this study incorporated a multifactor approach to measure the overall performance of the SFE process towards G. procumbens by using a rating and index approach. A summary of factors, such as the solubility of G. procumbens in CO2, operational cost and safety assessment elements, were taken into consideration as the main influences that determine the overall performance index of this study. Iperformance or overall performance of SFE from G. procumbens was successfully assessed and compared with response surface methodology (RSM). Overall, the results from Iperformance exhibit satisfactory solubility values when compared to the optimized value from RSM when considering the lowest operational costs in the safest SFE environment.
Supercritical CO2 extraction is usually applied to extract valuable compounds, including bioactive compounds from plant structures such as leaves, seeds, fruits and roots1,2,3. In Malaysia, the process of extracting herbs is rapidly developing. This was highly initiated in 2011 during the NKEA agriculture, and in one of the EPP projects, 18 types of herbs were chosen for further development. According to Dionysia 4, Gynura procumbens was the substance least used by traditional medicinal practitioners. G. procumbens, which is easily found in the tropical forests of Malaysia, Thailand and Indonesia, is an herb that contains useful compounds that can be used to relieve coughs, reduce blood glucose levels and prevent and treat liver cancer5. The herb is consumed raw as a salad or ‘ulam’ and can be applied topically. Moreover, G. procumbens contains flavonoids, saponins, tannins, and steroids, which all have potential as antioxidants6. The extracts contain medicinal ingredients, such as kaempferol 3-O-rutinoside, which can treat hypertension7; kaempferol, which is an anti-inflammatory8; and quercetin 3-o glucoside, which can treat diabetes9. Clinical studies were rigorously performed on the extracts, and all of the extracts were obtained by applying conventional and traditional techniques of extraction, such as solvent extraction using ethanol, methanol and water and hot and cold maceration techniques10. SC-CO2 extraction with ethanol–water has yet to be used in extracting valuable compounds from this herb.
The supercritical solvent used in SFE, which is CO2, is efficient in extracting non-polar components, such as terpenes and alkaloids, from plant samples11. Since the targeted compounds are antioxidants that are semi-polar, a co-solvent is introduced to improve the selectivity. For example, a semi-polar co-solvent, such as ethanol, was introduced to enhance the overall quality of the yield12. To date, many researchers have incorporated water into the co-solvent to further enhance the extraction process13,14,15,16. Water can modify the structure of the matrix inside the sample due to its higher viscosity compared to that of CO2 and ethanol and its lower solubility in CO2 compared to that of ethanol17.
Simulation of the extraction curve has been performed rigorously to predict the effect of parameters on the extract yields and to determine the optimum parameter for the best yield using various mathematical model approaches18. Sovová and Stateva19 reviewed types of mathematical models for SFE kinetics. According to them, there are five types of models, including the following: mass balance for solute, extraction steps and their characteristic times, a one-stage model, a model based on the complex structure of plant particles and a model for the SFE of mixtures19. Table 1 summarizes the reported study on applying a mathematical model in fitting the experimental data for the SFE-co-solvent process. The model was then analysed to determine the best operating conditions that produce the highest yield. According to Table 1, the pure ethanol and ethanol–water mixture was the most commonly used co-solvent for SFE. The broken and intact cell model (BIC) model was the most fitted model in fitting the experimental data. The BIC model is usually applied to a mechanically damaged cell sample due to the sample preparation20. The most important parameter for the BIC model was the initial fraction of easily accessible solute, G, in which a value between 0 and 1 was obtained21.
The experiment that obtained the highest yield under optimum conditions was also considered to achieve the best performance by the SFE process. The values of the optimum conditions can be evaluated by using a statistical tool, such as response surface methodology. This tool enables the user to choose the operating conditions, which can be optimized to obtain the highest yield. Moreover, the tool offers a good package that can provide a good design and analysis of the process by applying the statistical significance of all the factors used with analysis of variance. In addition, an artificial neural network is another piece of computing system that has been fully utilized in simulating the results of SFE. The network simulates the results by following the way the human brain analyses and processes information. This software is highly used due to its advantages, including that no thermophysical understanding of the SFE process is needed to conduct the simulation. In addition, previous knowledge about the neural network is not needed35. The ANN structure consists of a multi-layer, fully connected input layer, hidden layer, and output layer. The sensory data (experimental data) fed to the network is interpreted by the machine perceptron, which labels the input data and identifies the numerical patterns.
Table 2 shows the studies that have applied techniques to achieve the highest yield for the SFE process. Several studies have reported the application of two types of extraction techniques to achieve high yields in contrast to the process with SFE alone. For example, when extracting Caryocar Brasiliense, clove buds and Dipteryx alata, cold pressing was used together with SFE to achieve higher yield than that of SFE alone36,37,38. Economic evaluation has been conducted previously, in which the cost of manufacturing (COM) was determined. The calculated COM was compared between the SFE plant at the laboratory scale, pilot scale and industrial scale for production37,38,39. There is also a study focusing on economic assessment to evaluate the feasibility of the SFE process for the purpose of scale-up17. Moreover, there were also assessments of the safety in conducting the SFE process40. Most of these assessments were done separately and independently. None of the evaluations were systematically integrated to measure the overall performance of the SFE process.
To have an idea of the overall performance of SFE, a statistical report to measure the performance is needed. This study tends to consider the incorporation of multiple factors, such as yield, economic factors, and safety, in evaluating the overall performance of the SFE process of G. procumbens during the operation of the system. The results will be represented as an index that will act as an indicator to determine the overall performance of the SFE process.
Gynura procumbens was obtained from a local company, HERBagus Trading Sdn. Bhd which is located in Kepala Batas, Pulau Pinang, Malaysia. The sample received was cleaned and dried in open air for two days followed by drying using an oven at 50 °C until the total moisture content in the sample was less than 10%. The sample was then sieved to obtain a size of 2.0 mm and was stored at room temperature until use. The chemicals used were carbon dioxide (99.5% purity), which was purchased from Alpha Gas (Malaysia), and ethanol (99.8% purity), which was purchased from QRec (Malaysia).
A laboratory scale SFE unit that was designed and installed by a previous researcher was used for this study34. Chiller was first switched on to let the CO2 cool to − 4 °C to let the gas change to the liquid phase before being pumped to the SFE system. Three grams of G. procumbens ground leaves was inserted into the pressure vessel (H/D = 8). The oven was switched on, and the temperature was set to the designated operating temperature. The pressure on the back-pressure regulator was also set to its designated operating pressure. The combination factors for the SFE run for this study are shown in Table 3. The design of this experiment was performed by using a response surface method by Design Expert® (Stat-Ease, USA). Central composite design was utilized with three factors, which were the temperature (°C), pressure (MPa) and water content in ethanol (%). The response was fixed to the overall yield (g/g %) and solubility of G. procumbens in CO2 (g/g %). The α value was chosen as k > 5 with a value of 1.32. There are 20 runs, and the CCD positions of 8 factorial points, 6 axial points and 6 centre points are shown in Table 3.
To evaluate the economic element in the SFE of G. procumbens, the methodology by Turton et al. 45 was referred to when estimating the operational costs. Initially, they presented 3 categories to estimate the cost of manufacturing (COM), including the following: direct costs, fixed costs, and general expenses. Since this study focuses on calculating the operational cost, which is tabulated in Table 4, the fixed costs were not taken into consideration. The cost of waste treatment was also excluded since solid waste from SFE can be added to soil for the decomposition process. Therefore, the operational cost consists of the cost of raw materials (CRM), cost of utilities (CUT) and labour cost (COL). The economic parameter that was used to estimate the operational costs (OC) is also shown in Table 4. Therefore, the estimation of OC can be simplified from Turton et al. 45 to Eq. (1) as follows:
where OC is in units of RM/year.
Table 4 shows the description of each category in direct costs. In the raw materials costs (CRM), the price of CO2 dominates the costs. For utility costs (CUT), the use of electricity mostly originates from the equipment in the SFE system, as listed in Table 4.
First stage of the safety assessment.
Two objectives in evaluating the first stage were used to identify the hazard when conducting the SFE G. procumbens experiments and to classify the risk of hazards that can occur (light, moderate, intermediate, heavy, and severe). An analysis of the most hazardous equipment for SFE was conducted. After that, the potential of the second effect from the main scenario analysis was performed.
Second stage of safety assessment
A methodology by a previous researcher was used when evaluating the quantitative analysis of safety from SFE of G. procumbens46,47. This study used a mixture of ethanol–water as the co-solvent for the SFE process at a ratio of 10–30% v/v water–ethanol. Therefore, the risk of using the mixed co-solvent was estimated by identifying the boiling point (tb), flash point (fp) and Hansen solubility value (δ) for each of the ratios of water in ethanol. Equation (2) was applied to estimate tb. Equation (3) was used to determine the fp.
Equation (2) was used to determine the value of tb for the mixture. x1 is the mol fraction of solvent 1, x2 is the mol fraction of solvent 2, tb1 is the boiling point for solvent 1 and dan tb2 is the boiling point for solvent 2.
To estimate the flash point of the mixture, Eq. (3) was used.
where xi is the mol fraction, ({gamma }_{i}) is the activity coefficient, ({P}_{i,sat}) is the vapour pressure at T and ({P}_{i,sat FP}) is the vapour pressure at the flash point.
Equation (4) was used to determine the Hansen solubility value, whereby x1 is the mol fraction of solvent 1, x2 is the mol fraction of solvent 2, D1 is the solubility for solvent 1 and D2 is the solubility for solvent 2.
To estimate the Chemical Safety Total Score (CSTS), several factors need to be taken into consideration, and these factors are listed in Table 5. The factors that need to be determined are flammability, toxicity, reactivity, and explosiveness parameter. The equations used are displayed in Table 6. Then, the parameter was summed up as in Eq. (9) to obtain the value of CSTS.
Equation (10) was used to determine the performance index of SFE as follows:
All of the parameters were a function of temperature (T), pressure (P) and water content in ethanol ((omega)) and can be written as (fleft(T,P,omega right)). The solubility data were taken from RSM. The data for cost can be calculated based on Eq. (1), and safety was obtained from Eq. (9).
Both responses were successfully simulated using RSM, with yield regression using a 2-factor interaction (2FI) and solubility regression by a quadratic model. Table 6 shows the ANOVA by the CCD design and the details of the significant factor and coefficient for each factor for the regression equation.
From Table 6, Prob > F values are significant for both responses, and the values obtained were less than 0.05. Both responses were affected by the individual factors A and C and the interaction factor of A and C. The R2 values above 0.86 for both responses are also reasonable. Therefore, the reported research on applying CCD for the model regression for SFE was also indicated to be reasonable for use in this study48.
Table 7 shows the yield and solubility results from the experiment. Figures 1 and 2 show the response surface plot for the effect of temperature and pressure on the yield and solubility. Figure 1 shows that at 60 °C, the yield increased with pressure and improved at a higher temperature of 70 °C. This shows that the contact between the solute and solvent is better with pressure. The density of CO2 depends on the pressure and greatly affects the solubility of the solute in the CO249. Figure 2 clearly shows that the contour density is greatly affected by pressure compared to temperature for the solubility of G. procumbens in CO2. In other research, it was reported that when higher pressures are used for SFE, the effect of temperature on density is less noticeable, and the dominant factor is the vapour pressure50. When lower pressures are used, the change in temperature is more pronounced, and the process is dominated by the solvent density51. From here, we can see that there are variations in the solvation power of supercritical CO2 under different operating conditions.
Response surface plot for yield versus temperature and pressure.
Response surface plot for solubility versus temperature and pressure.
Figures 3 and 4 show the response surface plot for the effect of water content in ethanol on the yield and solubility at different pressures and water contents in ethanol. The effects of water content in ethanol can clearly be seen at the highest pressure of this experiment, which is 24 MPa. This is because of the enhancement of the solute solubility to the solvent, and this enhancement was influenced by the amount of water inside the ethanol14. When in contact with the sample, water can alter the sample matrix. A previous study reported that, compared to ethanol, water can better penetrate through the cell wall52. Water can also extract more lignin compounds in the secondary cell wall than in the layer between the cells52. This is due to the higher density of water compared to ethanol (Table 8). When this happens, the hole at the surface of the wall opens widely, causing the amount of lignin inside the sample to decrease. Moreover, a report from52 mentioned that carbon dioxide created an acidic environment when reacting with water. This triggers hemicellulose and lignin degradation on the primer cell wall. Therefore, the cell wall is no longer intact because the primer cell wall has been destroyed. Resistance towards the surface tension is also zero. Therefore, more CO2 can penetrate inside the cell to extract solute located at the secondary cell wall.
Response surface plot for yield versus pressure and water content in ethanol.
Response surface plot for solubility versus pressure and water content in ethanol.
Figure 3 shows that the highest extract was obtained at 24 MPa and was greater when the water content was increased from 10 to 30% inside ethanol. At low pressure, the water content does not have any effect on the yield obtained. Figure 4 clearly shows that the solubility value does not increase with increasing water content in ethanol. However, effective extraction occurred at the highest pressure of 24 MPa.
Figure 5 shows the fractions of the operational costs for the SFE of G. procumbens. The highest cost was the utility cost, which was 58% of the total, followed by labour cost (22%) and raw materials cost (20%). The highest contributor to the utility cost came from the electricity cost of the chiller, which utilized 13.82 kW, whereas other equipment, such as pumps, ovens and back-pressure regulators, contributed less than 1 kW each. Attard et al. (2015) reported that the same findings that CUT is the majority of the operational cost of conducting SFE53. The COM was further analysed to determine the distribution of each element (Fig. 6). The price of CO2 dominates the cost of CRM, with nearly 80%, followed by ethanol (20%). Previous research reported that the highest distributor in CRM was the cost of the sample. This is due to the sample supplier charging a high price. Moreover, a previous study was concerned with the higher production rate; therefore, a sample with a high mass was needed for extraction, resulting in higher cost54. The total operational costs were calculated for each parameter of the SFE G. procumbens. Then, it was further ranked to give the Icost value.
Individual factors in the CSTS score at each parameter for the SFE of G. procumbens.
Overall performance index, Iperformance, for SFE G. procumbens.
The main equipment in the SFE process that involve risks and hazards are listed in Table 9. Three pieces of equipment with the potential of experiencing overpressure were the CO2 pump, co-solvent pump, and pressure vessel. However, the following pieces of equipment involve a risk of boiling liquid expanding vapour exploration (BLEVE): CO2 storage tanks and pressure vessels. The types of hazards for each of the main equipment are also listed in the table. In this study, chemical hazards can occur with all the main equipment. This is because each of the equipment involves the solvent, which is the chemical in this process.
Table 10 shows the potential of secondary scenarios after the main scenario that can happen due to BLEVE. BLEVE is an explosion due to the failure of the pressure vessel, which is filled with liquid and is unable to withstand a temperature greater than the boiling point temperature at atmospheric pressure. According to Table 10, when overpressure occurs, it acts as a vector for the main scenario, after which a secondary potential scenario can form, such as a flash fire and toxic release. According to a previous study, studies on BLEVE were mostly performed on LPG and propane55. The effect of CO2 is still under review and can be further explored56.
The second safety assessment was conducted to determine the chemical safety total score (CSTS) by calculating each of the factors. According to Fig. 5, the CSTS scores were found to be the highest at 7% water content in ethanol. This is because the volume of ethanol was the largest. The highest portion for each CSTS score originated from the SFL score. This shows that the water content in ethanol does influence the flammability factor of the SFE process. SEXP and STX do not exhibit any variance at different parameters. However, SR has a small influence on CSTS. The CSTS value was then ranked to determine the safety index for this study.
The overall performance index for this study was determined by using Eq. (10). Individual indices including solubility, cost and safety were summed to obtain the total index, Iperformance. Figure 6 shows the overall performance of SFE G. procumbens at each parameter of the study. In addition, the individual index value was also shown to illustrate the factor that distributed the most for each performance. To choose the best performance from the parameters listed, the aim was to obtain the highest solubility at the lowest operational cost and in the safest environment. From Fig. 6, it can be concluded that the best Iperformance was obtained at 21 MPa, 65 °C and 33% water content in ethanol (v/v).
Table 11 shows the optimum results obtained from the RSM method and the best performance from the Iperformance method. Different parameter values were obtained from both. This was because the assessment for RSM does not consider the operating cost and safety factor when determining the optimum conditions, in which Iperformance integrates multiple factors to determine the best value. Overall, the results from Iperformance exhibit satisfactory solubility values when compared to the optimized value from RSM when considering the lowest operational costs in the safest SFE environment.
The overall performance index method is satisfactory for evaluating the SFE operation with G. procumbens. The results show that the value of optimum solubility from RSM did not differ much from the value obtained by Iperformance. However, a different parameter was chosen, whereby the pressure and temperature were chosen at the centre point and the water content in ethanol was selected at 33% (v/v) for the method by Iperformance. The water content affected the process as well as the safety of the SFE process, especially the flammability factor, SFL. Water, when added to ethanol, altered the matrix sample and assisted the mass transfer process of solute to the solvent (CO2 and ethanol). The economic evaluation reported that the highest cost in operational costs (OC) originated from utility costs (CUT), and the highest contributor was from the chiller. Breakdown of the raw materials costs (CRM) indicates that the cost of CO2 dominates the expense. The results from the safety assessments towards the SFE process imply that there were 2 types of risk that can occur to the pressure vessel, including BLEVE and overpressure. Moreover, a secondary potential scenario can occur when BLEVE is further boosted by overpressure. The solubility results from Iperformance are satisfactory compared to those from RSM. This suggests that the index method by rating of the individual factors of solubility, economy and safety was adequate to recommend the best operating conditions for the highest solubility, as well as for obtaining minimum operational costs and the safest conditions possible.
The datasets used and/or analysed during the current study available from the corresponding author on reasonable request.
Adeib, I. S., Norhuda, I., Roslina, R. N. & Ruzitah, M. S. Mass transfer and solubility of hibiscus cannabinus l. Seed oil in supercritical carbon dioxide. J. Appl. Sci. 10, (2010).
Idris, S. A., Markom, M., Abd Rahman, N. & Mohd Ali, J. Prediction of overall yield of Gynura procumbens from ethanol-water + supercritical CO2 extraction using artificial neural network model. Case Stud. Chem. Environ. Eng. 5, 100175 (2022).
CAS  Article  Google Scholar 
Zainuddin, N. A., Tuah, F. & Mohd Yatim, S. R. Supercritical carbon dioxide extraction of oil from Chromolaena odorata leaves. Heal. Scope 1, 152–156 (2019).
Google Scholar 
M Dionysia, MS Abdul Hayat, M Nik Musaadah, B Intan Nurulhani, MN Madihah, Z Nurul Husna, J Fadzureena, HF Lim, AL Tan, R Rosniza, MA Nor Azah, M Mastura & H Norini, Trend Penggunaan 18 Spesies Tumbuhan Ubatan Di Bawah Program Nkea Di Kalangan Pengamal Perubatan Tradisional Melayu Di Semenanjung Malaysia, Prosiding Persidangan Industri Herba, 7, 170–174 (2015).
Zhang, T., Gu, H.-W., Gao, J.-X., Li, Y.-S. & Tang, H.-B. Ethanol supernatant extracts of Gynura procumbens could treat nanodiethylnitrosamine-induced mouse liver cancer by interfering with inflammatory factors for the tumor microenvironment. J. Ethnopharmacol. 285, 114917 (2022).
CAS  PubMed  Article  Google Scholar 
Li, J. E., Wang, W. J., Zheng, G. D. & Li, L. Y. Physicochemical properties and antioxidant activities of polysaccharides from Gynura procumbens leaves by fractional precipitation. Int. J. Biol. Macromol. 95, 719–724 (2017).
CAS  PubMed  Article  Google Scholar 
Shahlehi, S. & Petalcorin, M. I. R. Activation of cholinergic pathway induced vasodilation in rat aorta using aqueous and methanolic leaf extracts of Gynura procumbens. Biomed. Pharmacother. 143, 112066 (2021).
CAS  PubMed  Article  Google Scholar 
Wong, S. K. et al. Anti-malarial and anti-inflammatory effects of gynura procumbens are mediated by kaempferol via inhibition of glycogen synthase kinase-3β (GSK3β). Sains Malaysiana 44, 1489–1500 (2015).
CAS  Article  Google Scholar 
Chan, C.-H., Yusoff, R., Ngoh, G. & Kung, F. W. Extraction of anti-diabetic active ingredient, quercetin from herbal plant using microwave-assisted extraction (MAE) Technique. in The International Conference on Materials for Advanced Technologies 2–5 (SUNTEC Singapore, 2011). https://doi.org/10.13140/2.1.3487.4885.
Ning, T. J. et al. Inhibitory effects of gynura procumbens ethanolic extract on nitric oxide production and inducible nitric oxide synthase (iNOS) protein expression in macrophages. Sains Malaysiana 48, 1737–1744 (2019).
CAS  Article  Google Scholar 
Guedes, A. R. et al. Extraction of Synadenium grantii Hook f. using conventional solvents and supercritical CO2 + ethanol. J. Supercrit. Fluids 160, 8–10 (2020).
Article  CAS  Google Scholar 
Radzali, S. A., Markom, M. & Saleh, N. M. Co-solvent selection for supercritical fluid extraction (SFE) of phenolic compounds from Labisia pumila. Molecules 25, 1–15 (2020).
Article  CAS  Google Scholar 
Hassim, N., Markom, M., Rosli, M. I. & Harun, S. Scale-up approach for supercritical fluid extraction with ethanol–water modified carbon dioxide on Phyllanthus niruri for safe enriched herbal extracts. Sci. Rep. 11, 1–19 (2021).
Article  CAS  Google Scholar 
Markom, M., Hassim, N., Hasan, M. & Daud, W. R. W. Modeling of supercritical fluid extraction by enhancement factor of cosolvent mixtures. Sep. Sci. Technol. 00, 1–13 (2020).
CAS  Google Scholar 
Idris, S. A. & Markom, M. Effect of water content in Co solvent on yield of supercritical fluid extraction of gynura procumbens leaves. J. Comput. Theor. Nanosci. 17, 1203–1206 (2020).
CAS  Article  Google Scholar 
Mohamed-Mahmood, M., Daud, W. R. W., Markom, M. & Mansor, C. N. A. N. C. Cosolvent selection for supercritical fluid extraction (SFE) of bioactive compounds from Orthosiphon stamineus. Sains Malaysiana 47, 1741–1747 (2018).
CAS  Article  Google Scholar 
Hassim, N., Markom, M., Rosli, M. I. & Harun, S. Scale-up criteria and economic analysis for supercritical fluid extraction of Phyllanthus niruri. Chem. Eng. Process. Process Intensif. 139, 14–22 (2019).
CAS  Article  Google Scholar 
Rai, A. Modeling Techniques in Empirical Supercritical Extraction Designs: Recent Trends and Practices. Innovative Food Processing Technologies Vol. 2 (Elsevier, 2021).
Google Scholar 
Sovovà, H. & Stateva, R. P. Supercritical fluid extraction from vegetable materials. Rev. Chem. Eng. 27, 79–156 (2011).
Article  CAS  Google Scholar 
Kupski, S. C. et al. Mathematical modeling of supercritical CO2 extraction of hops (Humulus lupulus L.). J. Supercrit. Fluids 130, 347–356 (2017).
CAS  Article  Google Scholar 
Aydi, A. et al. Supercritical CO2 extraction of extracted oil from Pistacia lentiscus L.: Mathematical modeling, economic evaluation and scale-up. Molecules 25, (2020).
Khodaie, F. & Ghoreishi, S. M. Experimental extraction of gallic acid from brown sumac seed (Rhus coriaria) using supercritical carbon dioxide and ethanol as co-solvent: Modeling and optimization. J. Supercrit. Fluids 175, 105266 (2021).
CAS  Article  Google Scholar 
Esquı́vel, M. M., Bernardo-Gil, M. G. & King, M. B. Mathematical models for supercritical extraction of olive husk oil. J. Supercrit. Fluids 16, 43–58 (1999).
Article  Google Scholar 
Martínez, J., Monteiro, A. R., Rosa, P. T. V., Marques, M. O. M. & Meireles, M. A. A. Multicomponent model to describe extraction of ginger oleoresin with supercritical carbon dioxide. Ind. Eng. Chem. Res. 42, 1057–1063 (2003).
Article  CAS  Google Scholar 
Tan, C. & Liou, D. Modeling of desorption at supercritical conditions. AIChE J. 35, 1029–1031 (1989).
CAS  Article  Google Scholar 
Sovová, H. Rate of the vegetable oil extraction with supercritical CO2—I. Modeling of extraction curves. Chem. Eng. Sci. 49, 409–414 (1994).
Article  Google Scholar 
Díaz-Reinoso, B., Moure, A. & Domínguez, H. Ethanol-modified supercritical Co2 extraction of chestnut burs antioxidants. Chem. Eng. Process. Process Intensif. 156, 108092 (2020).
Article  CAS  Google Scholar 
Grijó, D. R., Vieitez Osorio, I. A. & Cardozo-Filho, L. Supercritical extraction strategies using CO2 and ethanol to obtain cannabinoid compounds from Cannabis hybrid flowers. J. CO2 Util. 28, 174–180 (2018).
Article  CAS  Google Scholar 
Santos, K. A. et al. Candeia (Eremanthus erythroppapus) oil extraction using supercritical CO2 with ethanol and ethyl acetate cosolvents. J. Supercrit. Fluids 128, 323–330 (2017).
CAS  Article  Google Scholar 
del Valle, J. M., Martín, Á., Cocero, M. J., de la Fuente, J. C. & de la Cruz-Quiroz, R. Supercritical CO2 extraction of solids using aqueous ethanol as static modifier is a two-step mass transfer process. J. Supercrit. Fluids 143, 179–190 (2019).
Article  CAS  Google Scholar 
Sovová, H. Mathematical model for supercritical fluid extraction of natural products and extraction curve evaluation. J. Supercrit. Fluids 33, 35–52 (2005).
Article  CAS  Google Scholar 
Reverchon, E. Mathematical modeling of supercritical extraction of sage oil. AIChE J. 42, 1765–1771 (1996).
CAS  Article  Google Scholar 
Almeida, R. N. et al. Supercritical extraction of Hypericum caprifoliatum using carbon dioxide and ethanol+water as co-solvent. Chem. Eng. Process. Process Intensif. 70, 95–102 (2013).
CAS  Article  Google Scholar 
Markom, M. High-pressure extraction and fractionation of tannins from Phyllanthus Niruri Linn. (dukung Anak). (2007).
Pavlic, B., Bera, O., Vidovic, S., Ilic, L. & Zekovic, Z. Extraction kinetics and ANN simulation of supercritical fluid extraction of sage herbal dust. J. Supercrit. Fluids 130, 327–336 (2017).
CAS  Article  Google Scholar 
Johner, J. C. F., Hatami, T. & Meireles, M. A. A. Developing a supercritical fluid extraction method assisted by cold pressing for extraction of pequi (Caryocar brasiliense). J. Supercrit. Fluids 137, 34–39 (2018).
CAS  Article  Google Scholar 
Hatami, T., Johner, J. C. F., Zabot, G. L. & Meireles, M. A. A. Supercritical fluid extraction assisted by cold pressing from clove buds: Extraction performance, volatile oil composition, and economic evaluation. J. Supercrit. Fluids 144, 39–47 (2019).
CAS  Article  Google Scholar 
Chañi-Paucar, L. O., Osorio-Tobón, J. F., Johner, J. C. F. & Meireles, M. A. A. A comparative and economic study of the extraction of oil from Baru (Dipteryx alata) seeds by supercritical CO2 with and without mechanical pressing. Heliyon 7, e05971 (2021).
PubMed  PubMed Central  Article  CAS  Google Scholar 
de Aguiar, A. C., Osorio-Tobón, J. F., Silva, L. P. S., Barbero, G. F. & Martínez, J. Economic analysis of oleoresin production from malagueta peppers (Capsicum frutescens) by supercritical fluid extraction. J. Supercrit. Fluids 133, 86–93 (2018).
Article  CAS  Google Scholar 
Horvat, G., Aladić, K. & Jokić, S. Supercritical CO2 extraction pilot plant design—Towards IoT integration. Teh. Vjesn. 24, 925–934 (2017).
Google Scholar 
Vaeli, N., Honarvar, B., Esfandiari, N. & Aboosadi, Z. A. A laboratory study on extracting active ingredients from scrophularia striata boiss using ultrasound-assisted supercritical fluid extraction technique. S. Afr. J. Chem. Eng. 35, 111–117 (2021).
Google Scholar 
Klein, E. J. et al. Techno-economical optimization of uvaia (Eugenia pyriformis) extraction using supercritical fluid technology. J. Supercrit. Fluids 174, 105239 (2021).
CAS  Article  Google Scholar 
Murias, M. S., del Valle, J. M. & Núñez, G. A. Mathematical simulation of heat and mass transfer during controlled depressurization of supercritical CO2 in extraction vessels. J. Supercrit. Fluids 122, 43–51 (2017).
CAS  Article  Google Scholar 
Zeng, J. & Yang, S. X. Optimal control of supercritical fluid extraction with a hybrid model. in Proceedings of the 2003 IEEE International Symposium on Intelligent Control (2003).
Turton, R., Bailie, R. C., Whiting, W. B. & Shaeiwitz, J. A. Analysis, Synthesis and Design of Chemical Processes, 2008, Pearson Education.
Ahmad, S. I., Hashim, H. & Hassim, M. H. Numerical Descriptive Inherent Safety Technique (NuDIST) for inherent safety assessment in petrochemical industry. Process Saf. Environ. Prot. 92, 379–389 (2014).
CAS  Article  Google Scholar 
Ahmad, S. I. et al. Solvent design and inherent safety assessment of solvent alternatives for palm oil recovery. J. Loss Prev. Process Ind. 65, 104120 (2020).
CAS  Article  Google Scholar 
Manjare, S. D. & Dhingra, K. Supercritical fluids in separation and purification: A review. Mater. Sci. Energy Technol. 2, 463–484 (2019).
Google Scholar 
Guan, M., Xu, X., Tang, X. & Li, Y. Optimization of supercritical CO2 extraction by response surface methodology, composition analysis and economic evaluation of bamboo green wax. J. Clean. Prod. 330, 129906 (2022).
CAS  Article  Google Scholar 
Rai, A., Bhargava, R. & Mohanty, B. Simulation of supercritical fluid extraction of essential oil from natural products. J. Appl. Res. Med. Aromat. Plants 5, 1–9 (2017).
Google Scholar 
García-Pérez, J. S. et al. Thermodynamics and statistical correlation between supercritical-CO2 fluid extraction and bioactivity profile of locally available Mexican plants extracts. J. Supercrit. Fluids 122, 27–34 (2017).
Article  CAS  Google Scholar 
Jiang, Y., Feng, Y., Lei, B. & Zhong, H. Impact mechanisms of supercritical CO2–ethanol–water on extraction behavior and chemical structure of eucalyptus lignin. Int. J. Biol. Macromol. 161, 1506–1515 (2020).
CAS  PubMed  Article  Google Scholar 
Attard, T. M., McElroy, C. R. & Hunt, A. J. Economic assessment of supercritical CO2 extraction of waxes as part of a maize stover biorefinery. Int. J. Mol. Sci. 16, 17546–17564 (2015).
PubMed  PubMed Central  Article  Google Scholar 
Chañi-Paucar, L. O., Johner, J. C. F., Zabot, G. L. & Meireles, M. A. A. Technical and economic evaluation of supercritical CO2 extraction of oil from sucupira branca seeds. J. Supercrit. Fluids 181, 105494 (2022).
Article  CAS  Google Scholar 
Davidy, A. CFD simulation and mitigation with boiling liquid expanding vapor explosion (BLEVE) caused by jet fire. ChemEngineering 3, 1–22 (2019).
CAS  Article  Google Scholar 
Shang, Z. et al. Experimental investigation of BLEVE in liquid CO2 phase-transition blasting for enhanced coalbed methane recovery. Fuel 292, 120283 (2021).
CAS  Article  Google Scholar 
Download references
This research work is financially supported by Universiti Kebangsaan Malaysia (GUP-2019-009). The authors would also like to thank the Ministry of Higher Education (MOHE) (FRGS/1/2017/TK02/UKM/02/4) and Universiti Teknologi MARA for financial support and for providing fantastic facilities.
School of Chemical Engineering, College of Engineering, Universiti Teknologi MARA, 40450, Shah Alam, Selangor, Malaysia
Sitinoor Adeib Idris
Chemical Engineering Program, Faculty of Engineering & Built Environment, Universiti Kebangsaan Malaysia, 43600, Bangi, Selangor, Malaysia
Masturah Markom, Norliza Abd. Rahman & Jarinah Mohd Ali
Research Centre for Sustainable Process Technology (CESPRO), Faculty of Engineering & Built Environment, Universiti Kebangsaan Malaysia, 43600, Bangi, Selangor, Malaysia
Masturah Markom, Norliza Abd. Rahman & Jarinah Mohd Ali
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
The manuscript has been constructed and written by S.A.I. under the main supervision and funding acquisition of M.M. with co-supervision by N.A.R. and J.M.A.
Correspondence to Masturah Markom.
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
Idris, S.A., Markom, M., Abd. Rahman, N. et al. Multifactor assessments to determine the overall performance of supercritical fluid extraction from Gynura procumbens essential oil. Sci Rep 12, 14293 (2022). https://doi.org/10.1038/s41598-022-16773-4
Download citation
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41598-022-16773-4
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

Remembering Captain N: The Game Master, Gaming's First Multiverse – TheGamer

Think the concept of the multiverse is a hot new thing? Think again.
You folks remember the first time Simon Belmont, Mega Man, and Pit all fought together? Oh, it wasn’t Smash Bros. It was something far weirder and far worse.
Get into the time machine, losers, because we’re going back to an era when nobody gave a shit about any of this. An age before style guides and game bibles locked down the look and feel of every IP in video games. A time when Simon Belmont could be a fussy narcissist and Mega Man was both green and shouted the word “Mega” in every sentence. Also, they referred to Pit as “Kid Icarus” because – again – nobody gave a shit.
Related: Cool Spot Tricked Me Into Playing A 7 Up Commercial
Welcome to Captain N: The Game Master, a show that ran from 1989 to 1991 and maybe nobody but me remembers.
This was a real Saturday Morning Cartoon, folks. In fact, you can still watch almost all of it on YouTube because I’m certain nobody cares. Whoever owns the IP is either dead or too scared to bring this show back because it might be too good.
I’m also going to make an uneducated swing and say it’s also gaming’s first multiverse.
And by “multiverse,” I don’t just mean team-up. It was a multiverse creatively called Videoland, which sounds less like an interconnected hub of game universes and more like a place you used to rent VHS tapes before your grandmother died. Still, if you want the primordial version of the mandatory crossover hell we now live in, this is it.
The story goes like this: The main character is a teenager named Kevin who’s sucked into his television, I guess by way of his Nintendo Entertainment System. It’s never really clear. Also, his dog goes with him because why not. Kevin finds himself inside the games, where the beautiful Princess Toadstool Zelda Lana asks Kevin to help them save all of Videoland with the help of Mega Man, Simon Belmont, and “Kid Icarus.”
Who is Princess Lana? Nobody. She’s not from a game. I think they made her up so Kevin could have a romantic interest without them needing to insert an actual character. Why not Peach? She wasn’t Peach yet and there was a Super Mario cartoon. Why not Zelda? Because there was a Legend of Zelda cartoon and some – you won’t believe it – crossover episodes.
So they made up some lady. She’s royalty and runs Videoland while being passive-aggressively courted by Kevin and regular aggressively sexually harassed by Simon Belmont. I’m sorry, man. I know it’s a cartoon for kids, but it’s been over 30 years since the show was on the air and Simon Belmont would get his ass fired for the shit he tried to pull on that lady. There is something truly weird in turning the stolid, severe guy from the cover of Castlevania into a self-obsessed creep. You know what’s cooler than a vampire hunter? An unlikable vampire hunter.
Meanwhile, Kid Icarus (A.K.A. Pit) speaks in a broken Pig Latin-type dialect that borders on torture. He’s not as horrible a person as Simon Belmont, but you gotta buckle up. Full disclosure: I worked on Kid Icarus: Uprising and that game sounds nothing like this. That wasn’t a style choice; that was a not-ruining-the-game choice.
So, yes, nobody back then really cared what the show’s producers did with those characters. And obviously this show used its interconnected universes to market wholly unrelated games. On the other hand, what do you think gaming IP multiverses are now? They ain’t doing it to warm your hearts.
But here’s the rub: Captain N allowed a lot of weird shit. It didn’t always work, but at least it didn’t feel like they were painting within the lines set out by a corporate office. In later seasons, the characters were joined by a sentient Game Boy who could fly. They went on adventures with long-dead gaming heroes like Bayou Billy. There was an entire episode set in the world of Tetris, with all the people resembling blocks. And it was horrifying.
You want more? The main villain of the series was Mother Brain from Metroid. And if you know Mother Brain and Metroid canon, you know she is a loud, brassy lady who talks all the time. Was she voiced in this show by Levi Stubbs, the same guy who voiced the giant plant in the movie version of Little Shop of Horrors? Why, yes she was, thank you for asking.
Oh, and her henchmen were Eggplant Wizard from Kid Icarus and a blue version of King Hippo from Punch-Out. Because the ‘80s.
None of this really works that well. At all. There’s a solid reason the show didn’t last long. Even as a kid who loved this show, I kind of understood none of it made much sense. It feels like Captain N’s writers and producers were shown five minutes of each game and told to do whatever they wanted with the property. If that meant the heroes competed against Mega Man bosses in the Olympics, so be it. God bless.
Stakes in the show are all over the map, ranging from “Will Kevin ever see his family again?” to “Will Mega Man have a nice birthday?” Characters are often there to repeat the same information three or four times for stupid children like me. And the animation quality is… less than the best. Then again, it’s a Saturday Morning Cartoon.
And this is all coming from a fan. I watched every episode as a kid. I loved it. I still kind of love it. I literally – I’m not joking – have cel art from this show on my wall. Was it expensive? Well, getting it framed was! The art was under $10. The framing was over $100.
How do I know this art is official? Because it came with a certificate and, let’s be honest, nobody would ever fake it. It’s like counterfeiting a gift card to Borders. Sure, you could do it, but why would you?
At the same time, Captain N: The Game Master isn’t that different from the way multiverses are playing out in pop culture now. The MCU uses a multiverse to play with weirdo versions of their stock characters – which this show (accidentally) does. Warner Bros. uses a multiverse to have completely unrelated characters fight each other – which this show (intentionally) does. It’s not hard to imagine looking back in 30 years and thinking it was weird that Batman fought Shaggy. It’s not hard to imagine looking back now and thinking it’s weird that Batman fought Shaggy.
Captain N is wild. It’s not very good. But it’s wild.
Next: Wind Waker's Sprawling Ocean Is Still The Best Open World Ever Made
Mike Drucker is a television writer and standup comedian. He’s currently co-head writer and co-executive producer of Full Frontal with Samantha Bee, and has written for shows including The Tonight Show, Adam Ruins Everything, and Mystery Science Theater 3000. Mike’s writing has also appeared in video games for Sony and Nintendo. If you read this far, thank you.

source

How To Sidechain in FL Studio 20

Generalization in quantum machine learning from few training data – 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: 4919 (2022)
Metrics details
Modern quantum machine learning (QML) methods involve variationally optimizing a parameterized quantum circuit on a training data set, and subsequently making predictions on a testing data set (i.e., generalizing). In this work, we provide a comprehensive study of generalization performance in QML after training on a limited number N of training data points. We show that the generalization error of a quantum machine learning model with T trainable gates scales at worst as (sqrt{T/N}). When only K ≪ T gates have undergone substantial change in the optimization process, we prove that the generalization error improves to (sqrt{K/N}). Our results imply that the compiling of unitaries into a polynomial number of native gates, a crucial application for the quantum computing industry that typically uses exponential-size training data, can be sped up significantly. We also show that classification of quantum states across a phase transition with a quantum convolutional neural network requires only a very small training data set. Other potential applications include learning quantum error correcting codes or quantum dynamical simulation. Our work injects new hope into the field of QML, as good generalization is guaranteed from few training data.
The ultimate goal of machine learning (ML) is to make accurate predictions on unseen data. This is known as generalization, and significant effort has been expended to understand the generalization capabilities of classical ML models. For example, theoretical results have been formulated as upper bounds on the generalization error as a function of the training data size and the model complexity1,2,3,4,5. Such bounds provide guidance as to how much training data is required and/or sufficient to achieve accurate generalization.
Quantum machine learning (QML) is an emerging field that has generated great excitement6,7,8,9. Modern QML typically involves training a parameterized quantum circuit in order to analyze either classical or quantum data sets10,11,12,13,14,15,16. Early results indicate that, for classical data analysis, QML models may offer some advantage over classical models under certain circumstances17,18,19. It has also been proven that QML models can provide an exponential advantage in sample complexity for analyzing quantum data20,21.
However, little is known about the conditions needed for accurate generalization in QML. Significant progress has been made in understanding the trainability of QML models18,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36, but trainability is a separate question from generalization18,37,38. Overfitting of training data could be an issue for QML, just as it is for classical machine learning. Moreover, the training data size required for QML generalization has yet to be fully studied. Naïvely, one could expect that an exponential number of training points are needed when training a function acting on an exponentially large Hilbert space. For instance, some studies have found that, exponentially in n, the number of qubits, large amounts of training data would be needed, assuming that one is trying to train an arbitrary unitary39,40. This is a concerning result, since it would imply exponential scaling of the resources required for QML, which is precisely what the field of quantum computation would like to avoid.
In practice, a more relevant scenario to consider instead of arbitrary unitaries is learning a unitary that can be represented by a polynomial-depth quantum circuit. This class of unitaries corresponds to those that can be efficiently implemented on a quantum computer, and it is exponentially smaller than that of arbitrary unitaries. More generally, one could consider a QML model with T parameterized gates and relate the training data size N needed for generalization to T. Even more general would be to consider generalization error a dynamic quantity that varies during the optimization.
In this work, we prove highly general theoretical bounds on the generalization error in variational QML: The generalization error is approximately upper bounded by (sqrt{T/N}). In our proofs, we first establish covering number bounds for the class of quantum operations that a variational QML model can implement. From these, we then derive generalization error bounds using the chaining technique for random processes. A key implication of our results is that an efficiently implementable QML model, one such that (Tin {{{{{{{mathcal{O}}}}}}}}({{{{{{{rm{poly}}}}}}}}n)), only requires an efficient amount of training data, (Nin {{{{{{{mathcal{O}}}}}}}}({{{{{{{rm{poly}}}}}}}}n)), to obtain good generalization. This implication, by itself, will improve the efficiency guarantees of variational quantum algorithms10,41,42 that employ training data, such as quantum autoencoders13, quantum generative adversarial networks43, variational quantum error correction44,45, variational quantum compiling46,47, and variational dynamical simulation48,49,50,51. It also yields improved efficiency guarantees for classical algorithms that simulate QML models.
We furthermore refine our bounds to account for the optimization process. We show that generalization improves if only some parameters have undergone substantial change during the optimization. Hence, even if we used a number of parameters T larger than the training data size N, the QML model could still generalize well if only some of the parameters have changed significantly. This suggests that QML researchers should be careful not to overtrain their models especially when the decrease in training error is insufficient.
To showcase our results, we consider quantum convolutional neural networks (QCNNs)27,45, a QML model that has received significant attention. QCNNs have only (T={{{{{{{mathcal{O}}}}}}}}(log n)) parameters and yet they are capable of classifying quantum states into distinct phases. Our theory guarantees that QCNNs have good generalization error for quantum phase recognition with only polylogarithmic training resources, (Nin {{{{{{{mathcal{O}}}}}}}}({log }^{2}n)). We support this guarantee with a numerical demonstration, which suggests that even constant-size training data can suffice.
Finally, we highlight the task of quantum compiling, a crucial application for the quantum computing industry. State-of-the-art classical methods for approximate optimal compiling of unitaries often employ exponentially large training data sets52,53,54. However, our work indicates that only polynomial-sized data sets are needed, suggesting that state-of-the-art compilers could be further improved. Indeed, we numerically demonstrate the surprisingly low data cost of compiling the quantum Fourier transform at relatively large scales.
Let us first outline our theoretical framework. We consider a quantum machine learning model (QMLM) as being a parameterized quantum channel, i.e., a completely positive trace preserving (CPTP) map that is parameterized. We denote a QMLM as ({{{{{{{{mathcal{E}}}}}}}}}_{{{{{{{{boldsymbol{alpha }}}}}}}}}^{{{{{{{{rm{QMLM}}}}}}}}}(cdot )) where α = (θ, k) denotes the set of parameters, including continuous parameters θ inside gates, as well as discrete parameters k that allow the gate structure to vary. We make no further assumptions on the form of the dependence of the CPTP map ({{{{{{{{mathcal{E}}}}}}}}}_{{{{{{{{boldsymbol{alpha }}}}}}}}}^{{{{{{{{rm{QMLM}}}}}}}}}(cdot )) on the parameters α. During the training process, one would optimize the continuous parameters θ and potentially also the structure k of the QMLM.
A QMLM takes input data in the form of quantum states. For classical data x, the input is first encoded in a quantum state via a map x ↦ ρ(x). This allows the data to be either classical or quantum in nature, since regardless it is eventually encoded in a quantum state. We assume that the data encoding is fixed in advance and not optimized over. We remark here that our results also apply for more general encoding strategies involving data re-uploading55, as we explain in Supplementary Note 3.
For the sake of generality, we allow the QMLM to act on a subsystem of the state ρ(x). Hence, the output state can be written as (({{{{{{{{mathcal{E}}}}}}}}}_{{{{{{{{boldsymbol{alpha }}}}}}}}}^{{{{{{{{rm{QMLM}}}}}}}}}otimes {{{{{{mathrm{id}}}}}}})(rho (x))). For a given data point (xi, yi), we can write the loss function as
for some Hermitian observable ({O}_{{x}_{i},{y}_{i}}^{{{{{{{{rm{loss}}}}}}}}}). As is common in classical learning theory, the prediction error bounds will depend on the largest (absolute) value that the loss function can attain. In our case, we therefore assume ({C}_{{{{{{{{rm{loss}}}}}}}}}:=mathop{sup }nolimits_{x,y}|vert {O}_{x,y}^{{{{{{{{rm{loss}}}}}}}}}|vert < infty < ?A3B2 tal? > ), i.e., the spectral norm can be bounded uniformly over all possible loss observables.
In Eq. (1), we take the measurement to act on a single copy of the output of the QMLM ({{{{{{{{mathcal{E}}}}}}}}}_{{{{{{{{boldsymbol{alpha }}}}}}}}}^{{{{{{{{rm{QMLM}}}}}}}}}(cdot )) upon input of (a subsystem of) the data encoding state ρ(xi). At first this looks like a restriction. However, note that one can choose ({{{{{{{{mathcal{E}}}}}}}}}_{{{{{{{{boldsymbol{alpha }}}}}}}}}^{{{{{{{{rm{QMLM}}}}}}}}}(cdot )) to be a tensor product of multiple copies of a QMLM, each with the same parameter setting, applied to multiple copies of the input state. Hence our framework is general enough to allow for global measurements on multiple copies. In this addition to the aforementioned situation, we further study the case in which trainable gates are more generally reused.
For a training dataset (S={{({x}_{i},{y}_{i})}}_{i=1}^{N}) of size N, the average loss for parameters α on the training data is
which is often referred to as the training error. When we obtain a new input x, the prediction error of a parameter setting α is taken to be the expected loss
where the expectation is with respect to the distribution P from which the training examples are generated.
Achieving small prediction error R(α) is the ultimate goal of (quantum) machine learning. As P is generally not known, the training error ({hat{R}}_{S}({{{{{{{boldsymbol{alpha }}}}}}}})) is often taken as a proxy for R(α). This strategy can be justified via bounds on the generalization error
which is the key quantity that we bound in our theorems.
We prove probabilistic bounds on the generalization error of a QMLM. Our bounds guarantee that a good performance on a sufficiently large training data set implies, with high probability, a good performance on previously unseen data points. In particular, we provide a precise meaning of “sufficiently large” in terms of properties of the QMLM and the employed training procedure.
Figure 1 gives an overview of the different scenarios considered in this work. We begin with the basic form of our result. We consider a QMLM that has arbitrarily many non-trainable global quantum gates and T trainable local quantum gates. Here, by local we mean κ-local for some n-independent locality parameter κ, and a local quantum gate can be a unitary or a quantum channel acting on κ qubits. Then we have the following bound on the generalization error for the QMLM with final parameter setting α* after training:
Panel (a) depicts a basic QMLM with T = 6 independently parameterized gates. The gray boxes illustrate some global evolutions that are not trainable. Panel (b) shows a gate-sharing QMLM with T = 2 independently parameterized gates, each gate is repeatedly used for M = 3 times. In panel (c), we depict a multi-copy QMLM. We take measurement data from M rounds of a basic QMLM with T = 6 parameterized gates and post-process the measurement outcomes to produce an output. Running M copies of a basic QMLM with T gates is equivalent to running a gate-sharing QMLM with T = 6 parameterized gates, in which each gate is repeated M times. Panel (d) describes a gate-sharing QMLM under optimization. The parameterized gate to the left undergoes a small change, while the one to the right undergoes a large change. If we sort the changes Δ1, Δ2 from large to small, then Δ1 ≫ Δ2 ≈ 0. Finally, panel (e) illustrates gate-sharing QMLM with variable structure. The number T of parameterized gates changes throughout the optimization. The figure begins with T = 1 and ends with T = 2.
(Basic QMLM). For a QMLM with T parameterized local quantum channels, with high probability over training data of size N, we have that
Theorem 1 directly implies sample complexity bounds: For any ε > 0, we can, with high success probability, guarantee that gen(α*) ⩽ ε, already with training data of size (N sim Tlog T/{varepsilon }^{2}), which scales effectively linearly with T, the number of parameterized gates.
For efficiently implementable QMLMs with (Tin {{{{{{{mathcal{O}}}}}}}}({{{{{{{rm{poly}}}}}}}}n)), a sample size of (Nin {{{{{{{mathcal{O}}}}}}}}left({{{{{{{rm{poly}}}}}}}}n/{varepsilon }^{2}right)) is already sufficient. More concretely, if (Tin {{{{{{{mathcal{O}}}}}}}}left({n}^{D}right)) for some degree D, then the corresponding sufficient sample complexity obtained from Theorem 1 satisfies (Nin tilde{{{{{{{{mathcal{O}}}}}}}}}left({n}^{D}/{varepsilon }^{2}right)), where the (tilde{{{{{{{{mathcal{O}}}}}}}}}) hides factors logarithmic in n. In the NISQ era56, we expect the number T of trainable maps to only grow mildly with the number of qubits, e.g., as in the architectures discussed in refs. 18, 45, 57. In this case, Theorem 1 gives an especially strong guarantee.
In various QMLMs, such as QCNNs, the same parameterized local gates are applied repeatedly. One could also consider running the same QMLM multiple times to gather measurement data and then post-processing that data. In both cases, one should consider the QMLM as using the same parameterized local gates repeatedly. We assume each gate to be repeated at most M times. A direct application of Theorem 1 would suggest that we need a training data size N of roughly MT, the total number of parameterized gates. However, the required number of training data actually is much smaller:
(Gate-sharing QMLM). Consider a QMLM with T independently parameterized local quantum channels, where each channel is reused at most M times. With high probability over training data of size N, we have
Thus, good generalization, as in Remark 1, can already be guaranteed, with high probability, when the data size effectively scales linearly in T (the number of independently parameterized gates) and only logarithmically in M (the number of uses). In particular, applying multiple copies of the QMLM in parallel does not significantly worsen the generalization performance compared to a single copy. Thus, as we discuss in Supplementary Note 3, Theorem 2 ensures that we can increase the number of shots used to estimate expectation values at the QMLM output without substantially harming the generalization behavior.
The optimization process of the QMLM also plays an important role in the generalization performance. Suppose that during the optimization process, the tth local gate changed by a distance Δt. We can bound the generalization error by a function of the changes ({{{{{Delta }}}_{t}}}_{t}).
(Gate-sharing QMLM under optimization). Consider a QMLM with T independently parameterized local quantum channels, where the tth channel is reused at most M times and is changed by Δt during the optimization. Assume Δ1≥…≥ΔT. With high probability over training data of size N, we have
When only K ≪ T local quantum gates have undergone a significant change, then the generalization error will scale at worst linearly with K and logarithmically in the total number of parameterized gates MT. Given that recent numerical results suggest that the parameters in a deep parameterized quantum circuit only change by a small amount during training58,59, Theorem 3 may find application in studying the generalization behavior of deep QMLMs.
Finally, we consider a more advanced type of variable ansatz optimization strategy that is also adopted in practice60,61,62,63. Instead of fixing the structure of the QMLM, such as the number of parameterized gates and how the parameterized gates are interleaved with the fixed gates, the optimization algorithm could vary the structure, e.g., by adding or deleting parameterized gates. We assume that for each number T of parameterized gates, there are GT different QMLM architectures.
(Gate-sharing QMLM with variable structure). Consider a QMLM with an arbitrary number of parameterized local quantum channels, where for each T > 0, we have GT different QMLM architectures with T parameterized gates. Suppose that after optimizing on the data, the QMLM has T independently parameterized local quantum channels, each repeated at most M times. Then, with high probability over input training data of size N,
Thus, even if the QMLM can in principle use exponentially many parameterized gates, we can control the generalization error in terms of the number of parameterized gates used in the QMLM after optimization, and the dependence on the number of different architectures is only logarithmic. This logarithmic dependence is crucial as even in the cases when GT grows exponentially with T, we have (log ({G}_{T})/Nin {{{{{{{mathcal{O}}}}}}}}(T/N)).
In this section we present generalization error results obtained by simulating the following two QML implementations: (1) using a QCNN to classify states belonging to different quantum phases, and (2) training a parameterized quantum circuit to compile a quantum Fourier transform matrix.
We begin with the quantum phase classification application. The QCNN architecture introduced in45 generalizes the model of (classical) convolutional neural networks with the goal of performing pattern recognition on quantum data. It is composed of so-called convolutional and pooling layers, which alternate. In a convolutional layer, a sequence of translationally invariant parameterized unitaries on neighbouring qubits is applied in parallel, which works as a filter between feature maps in different layers of the QCNN. Then, in the pooling layers, a subset of the qubits are measured to reduce the dimensionality of the state while preserving the relevant features of the data. Conditioned on the corresponding measurement outcomes, translationally invariant parameterized 1-qubit unitaries are applied. The QCNN architecture has been employed for supervised QML tasks of classification of phases of matter and to devise quantum error correction schemes45. Moreover, QCNNs have been shown not to exhibit barren plateaus, making them a generically trainable QML architecture27.
The action of a QCNN can be considered as mapping an input state ρin to an output state ρout given as ({rho }_{{{mbox{out}}}}={{{{{{{{mathcal{E}}}}}}}}}_{{{{{{{{boldsymbol{alpha }}}}}}}}}^{{{{{{{{rm{QCNN}}}}}}}}}({rho }_{{{mbox{in}}}})). Then, given ρout, one measures the expectation value of a task-specific Hermitian operator.
In our implementation, we employ a QCNN to classify states belonging to different symmetry protected topological phases. Specifically, we consider the generalized cluster Hamiltonian
where Zi (Xi) denote the Pauli z (x) operator acting on qubit i, and where J1 and J2 are tunable coupling coefficients. As proved in64, and as schematically shown in Fig. 2, the ground-state phase diagram of the Hamiltonian of Eq. (9) has four different phases: symmetry-protected topological (I), ferromagnetic (II), anti-ferromagnetic (III), and trivial (IV). In the Methods section, we provide additional details regarding the classical simulation of the ground states of H.
We employed the QCNN architecture for quantum phase recognition on ground states of the generalized cluster Hamiltonian H of Eq. (9). We evaluated the phase assigned by the QCNN to a point in the J1–J2-plane by sampling 8192 computational basis measurement outcomes and taking the least frequent outcome as the predicted phase. Panel (a) visualizes the performance of the QCNN for 16-qubits, trained with 30 data points, which were labelled according to the analytically determined phase diagram. Blue crosses denote training data points (not all 30 are shown). Blue (red) circles represent correctly (incorrectly) classified points. Panel (b) shows that, as the training data size increases, the training accuracy quickly becomes a good predictor for the testing accuracy on 10,000 randomly sampled points, i.e., the dependence of testing accuracy on training accuracy is approximately linear with slope increasing with N. The different points in the plot correspond to different parameter settings in the QCNN throughout the optimization. The dotted gray line shows the baseline accuracy of 25% achieved by random guessing. Panel (c) shows that the improvement in the slope with growing training data size is similar for different numbers of qubits, reflecting the at-worst polylogarithmic dependence of N on n predicted by our bounds.
By sampling parameters in the (J1, J2) plane, we create a training set ({{(left|{psi }_{i}rightrangle,{y}_{i})}}_{i=1}^{N}) composed of ground states (left|{psi }_{i}rightrangle) of H and their associated labels yi. Here, the labels are in the form of length-two bit strings, i.e., yi ∈ {0, 1}2, where each possible bit string corresponds to a phase that (left|{psi }_{i}rightrangle) can belong to. The QCNN maps the n-qubit input state (left|{psi }_{i}rightrangle) to a 2-qubit output state. We think of the information about the phase as being encoded into the output state by which of the 4 computational basis effect operators is assigned the smallest probability. Namely, we define the loss function as (ell ({{{{{{{boldsymbol{alpha }}}}}}}};left|{psi }_{i}rightrangle,{y}_{i}):=leftlangle {y}_{i}right|{{{{{{{{mathcal{E}}}}}}}}}_{{{{{{{{boldsymbol{alpha }}}}}}}}}^{{{{{{{{rm{QCNN}}}}}}}}}(left|{psi }_{i}rightrangle leftlangle {psi }_{i}right|)left|, {y}_{i}rightrangle). This leads to an empirical risk given by
In Fig. 2, we visualize the phase classification performance achieved by our QCNN, trained according to this loss function, while additionally taking the number of misclassified points into account. Moreover, we show how the true risk, or rather the test accuracy as proxy for it, correlates well with the achieved training accuracy, already for small training data sizes. This is in agreement with our theoretical predictions, discussed in more detail in Supplementary Note 4, which for QCNNs gives a generalization error bound polylogarithmic in the number of qubits. We note that refs. 65, 66 observed similarly favorable training data requirements for a related task of learning phase diagrams.
Next, we turn our attention to the unitary compiling application. Compiling is the task of transforming a high-level algorithm into a low-level code that can be implemented on a device. Unitary compiling is a paradigmatic task in the NISQ era where a target unitary is compiled into a gate sequence that complies with NISQ device limitations, e.g., hardware-imposed connectivity and shallow depth to mitigate errors. Unitary compiling is crucial to the quantum computing industry, as it is essentially always performed prior to running an algorithm on a NISQ device, and various companies have their own commercial compilers67,68. Hence, any ability to accelerate unitary compiling could have industrial impact.
Here we consider the task of compiling the unitary U of the n-qubit Quantum Fourier Transform (QFT)69 into a short-depth parameterized quantum circuit V(α). For V(α) we employ the VAns (Variable Ansatz) algorithm62,70, which uses a machine learning protocol to iteratively grow a parameterized quantum circuit by placing and removing gates in a way that empirically leads to lower loss function values. Unlike traditional approaches that train just continuous parameters in a fixed structure circuit, VAns also trains discrete parameters, e.g., gate placement or type of gate, to explore the architecture hyperspace. In Supplementary Note 5, we apply our theoretical results in this compiling scenario.
The training set for compilation is of the form ({{left|{psi }_{i}rightrangle,Uleft|{psi }_{i}rightrangle }}_{i=1}^{N}), consisting of input states (left|{psi }_{i}rightrangle) and output states obtained through the action of U. The (left|{psi }_{i}rightrangle) are drawn independently from an underlying data-generating distribution. In our numerics, we consider three such distributions: (1) random computational basis states, (2) random (non-orthogonal) low-entangled states, and (3) Haar random n-qubit states. Note that states in the first two distributions are easy to prepare on a quantum computer, whereas states from the last distribution become costly to prepare as n grows. As the goal is to train V(α) to match the action of U on the training set, we define the loss function as the squared trace distance between (Uleft|{psi }_{i}rightrangle) and (V({{{{{{{boldsymbol{alpha }}}}}}}})left|{psi }_{i}rightrangle), i.e., (ell (alpha ;left|{psi }_{i}rightrangle,Uleft|{psi }_{i}rightrangle ):=vert|Uleft|{psi }_{i}rightrangle leftlangle {psi }_{i}right|{U}^{{{{dagger}}} }-V({{{{{{{boldsymbol{alpha }}}}}}}})left|{psi }_{i}rightrangle leftlangle {psi }_{i}right|V{({{{{{{{boldsymbol{alpha }}}}}}}})}^{{{{dagger}}} }|{|}_{1}^{2}). This leads to the empirical risk
where ∣∣ ⋅ ∣∣1 indicates the trace norm.
Figure 3 shows our numerical results. As predicted by our analytical results, we can, with high success probability, accurately compile the QFT when training on a data set of size polynomial in the number of qubits. Our numerical investigation shows a linear scaling of the training requirements when training on random computational basis states. This better than the quadratic scaling implied by a direct application of our theory, which holds for any arbitrary data-generating distribution. Approximate implementations of QFT with a reduced number of gates71, combined with our results, could help to further study this apparent gap theoretically. When training on Haar random states, our numerics suggest that an even smaller number of training data points is sufficient for good generalization: Up to n = 9 qubits, we generalize well from a constant number of training data points, independent of the system size.
We employed a variable structure QMLM (as discussed near Theorem 4). Panel (a) shows the dependence of ({N}_{min }), the minimum training data size for accurate compilation, on n, the number of qubits. Accurate compilation is defined as achieving (parallel U-V({{{{{{{boldsymbol{alpha }}}}}}}}){parallel }_{F}^{2} , < , 1{0}^{-5}) on 1 out of 8 (blue) or on 7 out of 8 (red) runs. For training data with random computational basis inputs (solid lines), ({N}_{min }) scales linearly in n. When training on examples with Haar random inputs (dashed lines), ({N}_{min }) is constant up to system size n = 9. In Panel (b), for n = 9 qubits, we plot the prediction error of successfully trained (training cost < 10−8) runs for 8 training data sets with N = 16 to N = 30 random computational basis inputs. Panel (c) shows the dependence of the testing error on the reciprocal of the training error for different training data sizes, in the case of 9 qubits. Here, the data consisted of random computational basis states and the corresponding outputs. As N increases, small training error becomes a more reliable predictor for small testing error. Each subplot shows 8 different training runs, trained on different training data sets.
Even more striking are our results when initializing close to the solution. In this case, as shown in Fig. 4, we find that two training data points suffice to obtain accurate generalization, which holds even up to a problem size of 40 qubits. Our theoretical results in Theorem 3 do predict reduced training requirements when initializing near the solution. Hence, the numerics are in agreement with the theory, although they paint an even more optimistic picture and suggest that further investigation is needed to understand why the training data requirements are so low. While the assumption of initialization near the solution is only viable assuming additional prior knowledge, it could be justified in certain scenarios. For example, if the unitaries to be compiled depend on a parameter, e.g., time, and if we have already compiled the unitary for one parameter setting, we might use this as initialization for unitaries with a similar parameter.
Each panel shows the results of a single randomly initialized training run, where we used randomly drawn low-entangled states for training. The testing error on 20 test states, which we allow to be more strongly entangled than the states used during training, is plotted versus the reciprocal of the training error for training data sizes N = 1, 2, for different system sizes n. A training data set of size N = 1 is not sufficient to guarantee good generalization: Even with decreasing training error, the testing error remains large. In contrast, assuming favorably initialized training, N = 2 training data points suffice for good generalization, even for up to n = 40 qubits.
We conclude by discussing the impact of our work on specific applications, a comparison to prior work, the interpretation of our results from the perspective of quantum advantage, and some open questions.
We begin with a discussion of the impact on specific applications. Quantum phase classification is an exciting application of QML, to which Ref. 45 has successfully applied QCNNs. However, Ref. 45 only provided a heuristic explanation for the good generalization performance of QCNNs. Here, we have presented a rigorous theory that encompasses QCNNs and explains their performance, and we have confirmed it numerically for a fairly complicated phase diagram and a wide range of system sizes. In particular, our analysis allows us to go beyond the specific model of QCNNs and extract general principles for how to ensure good generalization. As generating training data for this problem asks an experimenter to prepare a variety of states from different phases of matter, which will require careful tuning of different parameters in the underlying Hamiltonian, good generalization guarantees for small training data sizes are crucial to allow for the implementation of phase classification through QML in actual physical experiments.
Several successful protocols for unitary compiling make use of training data52,53,54. However, prior work has relied on training data sets whose size scaled exponentially with the number of qubits. This scaling is problematic, both because it suggests a similarly bad scaling of the computational complexity of processing the data and because generating training data can be expensive in actual physical implementations. Our generalization bounds provide theoretical guarantees on the performance that unitary compiling with only polynomial-size training data can achieve, for the relevant case of efficiently implementable target unitaries. As we have numerically demonstrated in the case of the Quantum Fourier Transform, this significant reduction in training data size makes unitary compiling scalable beyond what previous approaches could achieve. Moreover, our results provide new insight into why the VAns algorithm62 is successful for unitary compiling. We believe that the QML perspective on unitary compiling advocated for in this work will lead to new and improved ansätze, which could scale to even larger systems.
Recent methods for variational dynamical simulation rely on quantum compiling to compile a Trotterized unitary into a structured ansatz with the form of a diagonalization48,49,72,73. This technique allows for quantum simulations of times longer than an iterated Trotterization because parameters in the diagonalization may be changed by hand to provide longer-time simulations with a fixed depth circuit. We expect the quantum compiling results presented here to carry over to this application. This will allow these variational quantum simulation methods to use fewer training resources (either input-output pairs, or entangling auxiliary systems), yet still achieve good generalization and scalability.
Discovering quantum error correcting codes can be viewed as an optimization problem44,45,74,75,76,77,78. Furthermore, it can be thought of as a machine learning problem, since computing the average fidelity of the code involves training data (e.g., chosen from a 2-design44). Significant effort has been made to solve this problem on classical computers74,75,76,77,78. Such approaches can benefit from our generalization bounds, potentially leading to faster classical discovery of quantum codes. More recently, it was proposed to use near-term quantum computers to find such codes44,45. Again our bounds imply good generalization performance with small training data for this application, especially for QCNNs45, due to their logarithmic number of parameters.
Finally, autoencoders and generative adversarial networks (GANs) have recently been generalized to the quantum setting13,43,79,80. Both employ training data, and hence our generalization bounds provide quantitative guidance for how much training data to employ in these applications. Moreover, our results can provide guidance for ansatz design. While there is no standard ansatz yet for quantum autoencoders or quantum GANs, ansätze with a minimal number of parameters will likely lead to the best generalization performance.
Next, we give a comparison to previously known results. Some prior works have studied the generalization capabilities of quantum models, among them the classical learning-theoretic approaches of81,82,83,84,85,86,87,88,89; the more geometric perspective of17,18; and the information-theoretic technique of20,37. Independently of this work, Ref. 38 also investigated covering numbers in QMLMs. However our bounds are stronger, significantly more general, and broader in scope. We give a detailed comparison of our results to related work in Supplementary Note 1.
To view our results in the context of the quest for quantum advantage, it is important to note that we do not prove a quantum advantage of quantum over classical machine learning. However, generalization bounds for QMLMs are necessary to understand their potential for quantum advantage. Namely, QMLMs can outperform classical methods, assuming both achieve small training error, only in scenarios in which QMLMs generalize well, but classical ML methods do not. We therefore consider our results a guide in the search for quantum advantage of QML: We need to identify a task in which QMLMs with few trainable gates achieve small training error, but classical models need substantially higher model complexity to achieve the same goal. Then, our bounds guarantee that the QMLM performs well also on unseen data, but we expect the classical model to generalize poorly due to the high model complexity.
We conclude with some open questions. For QMLMs with exponentially many independently trainable gates, our generalization error bounds scale exponentially with n, and hence we do not make non-trivial claims about this regime. However, this does not yet imply that exponential-size QMLMs have bad generalization behavior. Whether and under which circumstances this is indeed the case is an interesting open question (e.g., see17,37 for some initial results). More generally, one can ask: Under what circumstances will a QMLM, even one of polynomial size, outperform our general bound. For example, if we have further prior knowledge about the loss, arising from specific target applications, it might be possible to use this information to tighten our generalization bounds. Moreover, as our generalization bounds are valid for arbitrary data-generating distributions, they may be overly pessimistic for favorable distributions. Concretely, in our numerical experiments for unitary compiling, highly entangled states were more favorable than especially efficiently preparable states from the perspective of generalization. It may thus be interesting to investigate distribution-specific tightenings of our results. Finally, it may be fruitful to combine the generalization bounds for QMLMs studied in this work and the effect of data encodings in86 to yield a better picture on generalization in quantum machine learning.
This section gives an overview over our techniques. First, we outline the proof strategy that leads to the different generalization bounds stated above. Second, we present more details about our numerical investigations.
An established approach to generalization bounds in classical statistical learning theory is to bound a complexity measure for the class under consideration. Metric entropies, i.e., logarithms of covering numbers, quantify complexity in exactly the way needed for generalization bounds, as one can show using the chaining technique from the theory of random processes90,91. Therefore, a high level view of our proof strategy is: We establish novel metric entropy bounds for QMLMs and then combine these with known generalization results from classical learning theory. The strongest form of our generalization bounds is the following.
(Mother theorem). Consider a QMLM with an arbitrary number of parameterized local quantum channels, where for each T > 0, we have GT different QMLM architectures with T trainable local gates. Suppose that after optimizing on the training data, the QMLM has T independently parameterized local quantum channels, where the tth channel is reused at most M times and is changed by Δt during the optimization. Without loss of generality, assume Δ1≥…≥ΔT. Then with high probability over input training data of size N, we have
where (f(K):=sqrt{frac{Klog (MT)}{N}}+mathop{sum }limits_{k=K+1}^{T}M{{{Delta }}}_{k}).
We give a detailed proof in Supplementary Note 3. There, we also describe a variant in case the loss function cannot be evaluated exactly, but only estimated statistically. Here, we present only a sketch of how to prove Theorem 5.
Before the proof sketch, however, we discuss how Theorem 5 relates to the generalization bounds stated above. In particular, we demonstrate how to obtain Theorems 1, 2, 3, and 4 as special cases of Theorem 5.
In the scenario of Theorem 1, the QMLM architecture is fixed in advance, each trainable map is only used once, and we do not take properties of the optimization procedure into account. In the language of Theorem 5, this means: There exists a single T > 0 with GT = 1 and we have ({G}_{tilde{T}}=0) for all (tilde{T}ne T). Also, M = 1. And instead of taking the minimum over K = 1, …, T, we consider the bound for K = T. Plugging this into the generalization bound of Theorem 5, we recover Theorem 1.
Similarly, Theorem 5 implies Theorems 2, 3, and 4. Namely, if we take GT = 1 and ({G}_{tilde{T}}=0) for all (tilde{T}ne T), and evaluate the bound for K = T, we recover Theorem 2. Choosing GT = 1 and ({G}_{tilde{T}}=0) for all (tilde{T}ne T), the bound of Theorem 5 becomes that of Theorem 3. Finally, we can obtain Theorem 4 by bounding the minimum in Theorem 5 in terms of the expression evaluated at K = T.
Now that we have established that Theorem 5 indeed implies generalization bounds for all the different scenarios depicted in Fig. 1, we outline its proof. The first central ingredient to our reasoning are metric entropy bounds for the class of all n-qubit CPTP maps that a QMLM as described in Theorem 5 can implement, where the distance between such maps is measured in terms of the diamond norm. Note: The trivial metric entropy bound obtained by considering this class of maps as compact subset of an Euclidean space of dimension exponential in n is not sufficient for our purposes since it scales exponentially in n. Instead, we exploit the layer structure of QMLMs to obtain a better bound. More precisely, we show: If we fix a QMLM architecture with T trainable 2-qubit maps and a number of maps 0 ⩽ K ⩽ T, and we assume (data-dependent) optimization distances Δ1⩾…⩾ΔT, then it suffices to take (ε/KM)-covering nets for each of the sets of admissible 2-qubit CPTP maps for the first K trainable maps to obtain a ((varepsilon+mathop{sum }nolimits_{k=K+1}^{T}M{{{Delta }}}_{k}))-covering net for the whole QMLM. The cardinality of a covering net built in this way, crucially, is independent of n, but depends instead on K, M, and T. In detail, its logarithm can effectively be bounded as (in {{{{{{{mathcal{O}}}}}}}}left(Klog left(MT/varepsilon right)right)). This argument directly extends from the 2-local to the κ-local case, as we describe in Supplementary Note 3.
Now we employ the second core ingredient of our proof strategy. Namely, we combine a known upper bound on the generalization error in terms of the expected supremum of a certain random process with the so-called chaining technique. This leads to a generalization error bound in terms of a metric entropy integral. As we need a non-standard version of this bound, we provide a complete derivation for this strengthened form. This then tells us that, for each fixed T, M, K, and Δ1⩾…⩾ΔT, using the covering net constructed above, we can bound the generalization error as ({{{{{{mathrm{gen}}}}}}},({{{{{{{{boldsymbol{alpha }}}}}}}}}{*})in {{{{{{{mathcal{O}}}}}}}}(sqrt{Klog (MT)/N}+mathop{sum }nolimits_{k=K+1}^{T}M{{{Delta }}}_{k})), with high probability.
The last step of the proof consists of two applications of the union bound. The first instance is a union bound over the possible values of K. This leads to a generalization error bound in which we minimize over K = 0, …, T. So far, however, the bound still applies only to any QMLM with fixed architecture. We extend it to variable QMLM architectures by taking a second union bound over all admissible numbers of trainable gates T and the corresponding GT architectures. As this is, in general, a union bound over countably many events, we have to ensure that the corresponding failure probabilities are summable. Thus, we invoke our fixed-architecture generalization error bound for a success probability that is proportional to ({({G}_{T}{T}^{2})}^{-1}). In that way, the union bound over all possible architectures yields the logarithmic dependence on GT in the final bound and completes the proof of Theorem 5.
This section discusses numerical methods used throughout the paper. The subsections give details on computational techniques applied to phase classification of the cluster Hamiltonian in Eq. (9) and Quantum Fourier Transform compilation.
The training and testing sets consist of ground states (left|{psi }_{i}rightrangle) of the cluster Hamiltonian in Eq. (9), computed for different coupling strengths (J1, J2). The states (left|{psi }_{i}rightrangle) were obtained with the translation invariant Density Matrix Renormalization Group92. The states in the training set (represented by blue crosses in Fig. 2a) are chosen to be away from phase transition lines, so accurate description of the ground states is already achieved at small bond dimension χ. That value determines the cost of further computation involving the states (left|{psi }_{i}rightrangle) and we keep it small for efficient simulation.
We use Matrix Product State techniques93 to compute and optimize the empirical risk in Eq. (10). The main part of that calculation is the simulation of the action of the QCNN ({{{{{{{{mathcal{E}}}}}}}}}_{{{{{{{{boldsymbol{alpha }}}}}}}}}^{{{{{{{{rm{QCNN}}}}}}}}}) on a given ground state (left|{psi }_{i}rightrangle). The map ({{{{{{{{mathcal{E}}}}}}}}}_{{{{{{{{boldsymbol{alpha }}}}}}}}}^{{{{{{{{rm{QCNN}}}}}}}}}) consists of alternating convolutional and pooling layers. In our implementation the layers are translationally invariant and are represented by parameterized two-qubit gates. The action of a convolutional layer on an MPS amounts to updating two nearest neighbor MPS tensors in a way similar to the time-evolving block decimation algorithm94. The pooling layer is simulated in two steps. First, we simulate the action of all two-qubit gates on an MPS. This is analogous to the action of a convolutional layer, but performed on a different pair of nearest neighbor MPS tensors. This step is followed by a measurement of half of the qubits. We use the fact that the MPS can be written as a unitary tensor network and hence allows for perfect sampling techniques95. The measurement step results in a reduction of the system size by a factor of two.
We repeat the application of convolutional and pooling layers using the MPS as described above until the system size becomes small enough to allow for an exact description. A few final layers are simulated in a standard way and the empirical risk is given by a two-qubit measurement according to the label yi, as in Eq. (10). The empirical risk is optimized with the Simultaneous Perturbation Stochastic Approximation algorithm96. We grow the number of shots used in pooling layer measurements as the empirical risk is minimized. This results in a shot-frugal optimization97, as one can control the accuracy of the gradient based on the current optimization landscape.
In the Numerical results section, we show that the task of unitary compilation can be translated into minimization of the empirical risk ({hat{R}}_{S}({{{{{{{boldsymbol{alpha }}}}}}}})) defined in Eq. (11). Here, α = (θ, k) denotes a set of parameters that specifies a trainable unitary V(α). The optimization is performed in the space of all shallow circuits. It has discrete and continuous components. The discrete parameters k control the circuit layout, that is, the placement of all gates used in the circuit. Those gates are described by the continuous parameters θ. The optimization (mathop{min }limits_{{{{{{{{boldsymbol{alpha }}}}}}}}}{hat{R}}_{S}({{{{{{{boldsymbol{alpha }}}}}}}})) is performed with the recently introduced VAns algorithm62,70. The unitary V(α) is initialized with a circuit that consists of a few randomly placed gates. In subsequent iterations, VAns modifies the structure parameter k according to certain rules that involve randomly placing a resolution of the identity and removing gates that do not significantly contribute to the minimization of the empirical risk ({hat{R}}_{S}({{{{{{{boldsymbol{alpha }}}}}}}})). The qFactor algorithm54, modified to work with a set of pairs of states as opposed to a target unitary, is used to optimize over continuous parameters θ for fixed k. This optimization is performed after each update to the structure parameter k. In subsequent iterations, VAns makes a probabilistic decision whether the new set of parameters ({{{{{{{boldsymbol{{alpha }}}}}}}^{prime}}}) is kept or rejected. This decision is based on the change in empirical risk ({hat{R}}_{S}({{{{{{{boldsymbol{{alpha }}}}}}}^{prime}}})-{hat{R}}_{S}({{{{{{{boldsymbol{alpha }}}}}}}})), an artificial temperature T, and a factor Λ that sets the penalty for growing the circuit too quickly. To that end, we employ a simulated annealing technique, gradually decreasing T and Λ, and repeat the iterations described above until ({hat{R}}_{S}({{{{{{{boldsymbol{alpha }}}}}}}})) reaches a sufficiently small value.
Let us now discuss the methods used to optimize the empirical risk when V(α) is initialized close to the solution. Here, we start with a textbook circuit for performing the QFT and modify it in the following way. First, the circuit is rewritten such that it consists of two-qubit gates only. Next, each two-qubit gate u is replaced with (u^{prime}=u{e}^{idelta h}), where h is a random Hermitian matrix and δ is chosen such that (|vert u-u^{prime}|vert=epsilon) for an initially specified ϵ. The results presented in the Numerical results section are obtained with ϵ = 0.1. The perturbation considered here does not affect the circuit layout and hence the optimization over continuous parameters θ is sufficient to minimize the empirical risk ({hat{R}}_{S}({{{{{{{boldsymbol{alpha }}}}}}}})). We use qFactor to perform that optimization.
The input states (left|{psi }_{i}rightrangle) in the training set ({{left|{psi }_{i}rightrangle,{U}_{{{{{{{{rm{QFT}}}}}}}}}left|{psi }_{i}rightrangle }}_{i=1}^{N}) are random MPSs of bond dimension χ = 2. The QFT is efficiently simulable98 for such input states, which means that ({U}_{{{{{{{{rm{QFT}}}}}}}}}left|{psi }_{i}rightrangle) admits an efficient MPS description. Indeed, we find that a bond dimension χ < 20 is sufficient to accurately describe ({U}_{{{{{{{{rm{QFT}}}}}}}}}left|{psi }_{i}rightrangle). In summary, the use of MPS techniques allows us to construct the training set efficiently. Note that the states (V({{{{{{{boldsymbol{alpha }}}}}}}})left|{psi }_{i}rightrangle) are in general more entangled than ({U}_{{{{{{{{rm{QFT}}}}}}}}}left|{psi }_{i}rightrangle), especially at the beginning of the optimization. Because of that, we truncate the evolved MPS during the optimization. We find that a maximal allowed bond dimension of 100 is large enough to perform stable, successful minimization of the empirical risk with qFactor. The testing is performed with 20 randomly chosen initial states. We test with bond dimension χ = 10 MPSs, so the testing is done with more strongly entangled states than the training. Additionally, for system sizes up to 16 qubits, we verify that the trained unitary V is close (in the trace norm) to UQFT, when training is performed with at least two states.
The data generated and analyzed during the current study are available from the authors upon request.
Further implementation details are available from the authors upon request.
Vapnik, V. N. & Chervonenkis, A. Y. On the uniform convergence of relative frequencies of events to their probabilities. Th. Prob. App. 16, 264–280 (1971).
MATH  Article  Google Scholar 
Pollard, D. Convergence of stochastic processes (Springer, 1984).
Giné, E. & Zinn, J. Some limit theorems for empirical processes. Ann. Probability 929–989. https://doi.org/10.1214/aop/1176993138 (1984).
Dudley, R. M. Uniform Central Limit Theorems (Cambridge University Press, 1999).
Bartlett, P. L. & Mendelson, S. Rademacher and gaussian complexities: Risk bounds and structural results. J. Machine Learning Res. 3, 463–482 (2002).
MathSciNet  MATH  Google Scholar 
Biamonte, J. Quantum machine learning. Nature 549, 195–202 (2017).
ADS  CAS  PubMed  Article  Google Scholar 
Schuld, M., Sinayskiy, I. & Petruccione, F. An introduction to quantum machine learning. Contemporary Phys. 56, 172–185 (2015).
ADS  MATH  Article  Google Scholar 
Schuld, M., Sinayskiy, I. & Petruccione, F. The quest for a quantum neural network. Quantum Inf. Process. 13, 2567–2586 (2014).
ADS  MathSciNet  MATH  Article  Google Scholar 
Dunjko, V. & Briegel, H. J. Machine learning & artificial intelligence in the quantum domain: a review of recent progress. Rep. Prog. Phys. 81, 074001 (2018).
ADS  MathSciNet  PubMed  Article  Google Scholar 
Cerezo, M. Variational quantum algorithms. Nat. Rev. Phys. 3, 625–644 (2021).
Article  Google Scholar 
Havlíček, V. Supervised learning with quantum-enhanced feature spaces. Nature 567, 209–212 (2019).
ADS  PubMed  Article  CAS  Google Scholar 
Farhi, E. & Neven, H. Classification with quantum neural networks on near term processors, arXiv preprint arXiv:1802.06002 (2018).
Romero, J., Olson, J. P. & Aspuru-Guzik, A. Quantum autoencoders for efficient compression of quantum data. Quantum Sci. Technol. 2, 045001 (2017).
ADS  Article  Google Scholar 
Wan, K. H., Dahlsten, O., Kristjánsson, H., Gardner, R. & Kim, M. S. Quantum generalisation of feedforward neural networks. npj Quantum Inf. 3, 1–8 (2017).
CAS  Article  Google Scholar 
Larocca, M., Ju, N., García-Martín, D., Coles, P. J. & Cerezo, M. Theory of overparametrization in quantum neural networks, arXiv preprint arXiv:2109.11676 (2021).
Schatzki, L., Arrasmith, A., Coles, P. J. & Cerezo, M. Entangled datasets for quantum machine learning, arXiv preprint arXiv:2109.03400 (2021).
Huang, H. Y. Power of data in quantum machine learning. Nat. Commun. 12, 1–9 (2021).
ADS  CAS  Article  Google Scholar 
Abbas, A. The power of quantum neural networks. Nat. Comput. Sci. 1, 403–409 (2021).
Article  Google Scholar 
Liu, Y., Arunachalam, S. & Temme, K. A rigorous and robust quantum speed-up in supervised machine learning, https://doi.org/10.1038/s41567-021-01287-zNat. Phys., 1–5 (2021).
Huang, H. Y., Kueng, R. & Preskill, J. Information-theoretic bounds on quantum advantage in machine learning. Phys. Rev. Lett. 126, 190505 (2021).
ADS  MathSciNet  CAS  PubMed  Article  Google Scholar 
Aharonov, D., Cotler, J. & Qi, X. L. Quantum algorithmic measurement. Nat. Commun. 13, 1–9 (2022).
Article  CAS  Google Scholar 
McClean, J. R., Boixo, S., Smelyanskiy, V. N., Babbush, R. & Neven, H. Barren plateaus in quantum neural network training landscapes. Nat. Commun. 9, 1–6 (2018).
ADS  CAS  Article  Google Scholar 
Cerezo, M., Sone, A., Volkoff, T., Cincio, L. & Coles, P. J. Cost function dependent barren plateaus in shallow parametrized quantum circuits. Nat. Commun. 12, 1–12 (2021).
Article  CAS  Google Scholar 
Cerezo, M. & Coles, P. J. Higher order derivatives of quantum neural networks with barren plateaus. Quantum Sci. Technol. 6, 035006 (2021).
ADS  Article  Google Scholar 
Arrasmith, A., Cerezo, M., Czarnik, P., Cincio, L. & Coles, P. J. Effect of barren plateaus on gradient-free optimization. Quantum 5, 558 (2021).
Article  Google Scholar 
Holmes, Z., Sharma, K., Cerezo, M. & Coles, P. J. Connecting ansatz expressibility to gradient magnitudes and barren plateaus. PRX Quantum 3, 010313 (2022).
ADS  Article  Google Scholar 
Pesah, A. Absence of barren plateaus in quantum convolutional neural networks. Phys. Rev. X 11, 041011 (2021).
CAS  Google Scholar 
Volkoff, T. & Coles, P. J. Large gradients via correlation in random parameterized quantum circuits. Quantum Sci. Technol. 6, 025008 (2021).
ADS  Article  Google Scholar 
Sharma, K., Cerezo, M., Cincio, L. & Coles, P. J. Trainability of dissipative perceptron-based quantum neural networks. Phys. Rev. Lett. 128, 180505 (2022).
ADS  MathSciNet  CAS  PubMed  Article  Google Scholar 
Holmes, Z. Barren plateaus preclude learning scramblers. Phys. Rev. Lett. 126, 190501 (2021).
ADS  MathSciNet  CAS  PubMed  Article  Google Scholar 
Marrero, C. O., Kieferová, M. & Wiebe, N. Entanglement-induced barren plateaus. PRX Quantum 2, 040316 (2021).
Article  Google Scholar 
Uvarov, A. V. & Biamonte, J. D. On barren plateaus and cost function locality in variational quantum algorithms. J. Phys. A: Math. Theor. 54, 245301 (2021).
ADS  MathSciNet  Article  Google Scholar 
Patti, T. L., Najafi, K., Gao, X. & Yelin, S. F. Entanglement devised barren plateau mitigation. Phys. Rev. Res. 3, 033090 (2021).
CAS  Article  Google Scholar 
Wang, S. Noise-induced barren plateaus in variational quantum algorithms. Nat. Commun. 12, 1–11 (2021).
Article  CAS  Google Scholar 
Larocca, M. et al. Diagnosing barren plateaus with tools from quantum optimal control, arXiv preprint arXiv:2105.14377 (2021).
Thanaslip, S., Wang, S., Nghiem, N. A., Coles, P. J. & Cerezo, M. Subtleties in the trainability of quantum machine learning models, arXiv preprint arXiv:2110.14753 (2021).
Banchi, L., Pereira, J. & Pirandola, S. Generalization in quantum machine learning: a quantum information standpoint. PRX Quantum 2, 040321 (2021).
ADS  Article  Google Scholar 
Du, Y., Tu, Z., Yuan, X. & Tao, D. Efficient measure for the expressivity of variational quantum algorithms. Phys. Rev. Lett. 128, 080506 (2022).
ADS  MathSciNet  CAS  PubMed  Article  Google Scholar 
Poland, K., Beer, K. & Osborne, T. J. No free lunch for quantum machine learning, arXiv preprint arXiv:2003.14103 (2020).
Sharma, K. Reformulation of the no-free-lunch theorem for entangled datasets. Phys. Rev. Lett. 128, 070501 (2022).
ADS  MathSciNet  CAS  PubMed  Article  Google Scholar 
Bharti, K. Noisy intermediate-scale quantum algorithms. Rev. Modern Phys. 94, 015004 (2022).
ADS  MathSciNet  CAS  Article  Google Scholar 
Endo, S., Cai, Z., Benjamin, S. C. & Yuan, X. Hybrid quantum-classical algorithms and quantum error mitigation. J. Phys. Soc. Japan 90, 032001 (2021).
ADS  Article  Google Scholar 
Romero, J. & Aspuru-Guzik, A. Variational quantum generators: Generative adversarial quantum machine learning for continuous distributions. Adv. Quantum Technol. 4, 2000003 (2021).
Article  Google Scholar 
Johnson, P. D., Romero, J., Olson, J., Cao, Y. & Aspuru-Guzik, A. Qvector: an algorithm for device-tailored quantum error correction, arXiv preprint arXiv:1711.02249 (2017).
Cong, I., Choi, S. & Lukin, M. D. Quantum convolutional neural networks. Nat. Phys. 15, 1273–1278 (2019).
CAS  Article  Google Scholar 
Khatri, S. Quantum-assisted quantum compiling. Quantum 3, 140 (2019).
Article  Google Scholar 
Sharma, K., Khatri, S., Cerezo, M. & Coles, P. J. Noise resilience of variational quantum compiling. New J. Phys. 22, 043006 (2020).
ADS  MathSciNet  Article  Google Scholar 
Cirstoiu, C. Variational fast forwarding for quantum simulation beyond the coherence time. npj Quantum Inf. 6, 1–10 (2020).
Article  Google Scholar 
Commeau, B. et al. Variational hamiltonian diagonalization for dynamical quantum simulation, arXiv preprint arXiv:2009.02559 (2020).
Endo, S., Sun, J., Li, Y., Benjamin, S. C. & Yuan, X. Variational quantum simulation of general processes. Phys. Rev. Lett. 125, 010501 (2020).
ADS  MathSciNet  CAS  PubMed  Article  Google Scholar 
Li, Y. & Benjamin, S. C. Efficient variational quantum simulator incorporating active error minimization. Phys. Rev. X 7, 021050 (2017).
Google Scholar 
Cincio, L., Subaşí, Y., Sornborger, A. T. & Coles, P. J. Learning the quantum algorithm for state overlap. New J. Phys. 20, 113022 (2018).
ADS  Article  Google Scholar 
Cincio, L., Rudinger, K., Sarovar, M. & Coles, P. J. Machine learning of noise-resilient quantum circuits. PRX Quantum 2, 010324 (2021).
Article  Google Scholar 
Younis, E. & Cincio, L. https://github.com/BQSKit/qfactor Quantum Fast Circuit Optimizer (qFactor).
Pérez-Salinas, A., Cervera-Lierta, A., Gil-Fuster, E. & Latorre, J. Data re-uploading for a universal quantum classifier. Quantum 4, 226 (2020).
Article  Google Scholar 
Preskill, J. Quantum computing in the nisq era and beyond. Quantum 2, 79 (2018).
Article  Google Scholar 
Romero, J. Strategies for quantum computing molecular energies using the unitary coupled cluster ansatz. Quantum Sci. Technol. 4, 014008 (2018).
ADS  Article  Google Scholar 
Shirai, N., Kubo, K., Mitarai, K. & Fuji, K. Quantum tangent kernel, arXiv preprint arXiv:2111.02951 (2021).
Liu, J., Tacchino, F., Glick, J. R., Jiang, L. & Mezzacapo, A. Representation learning via quantum neural tangent kernels, arXiv preprint arXiv:2111.04225 (2021).
Grimsley, H. R., Economou, S. E., Barnes, E. & Mayhall, N. J. An adaptive variational algorithm for exact molecular simulations on a quantum computer. Nat. Commun. 10, 1–9 (2019).
CAS  Article  Google Scholar 
Tang, H. L. qubit-adapt-vqe: an adaptive algorithm for constructing hardware-efficient ansätze on a quantum processor. PRX Quantum 2, 020310 (2021).
ADS  Article  Google Scholar 
Bilkis, M., Cerezo, M., Verdon, G., Coles, P. J. & Cincio, L. A semi-agnostic ansatz with variable structure for quantum machine learning, arXiv preprint arXiv:2103.06712 (2021).
Zhu, L. et al. An adaptive quantum approximate optimization algorithm for solving combinatorial problems on a quantum computer, arXiv preprint arXiv:2005.10258 (2020).
Verresen, R., Moessner, R. & Pollmann, F. One-dimensional symmetry protected topological phases and their transitions. Phys. Rev. B 96, 165124 (2017).
ADS  Article  Google Scholar 
Kottmann, K., Corboz, P., Lewenstein, M. & Acín, A. Unsupervised mapping of phase diagrams of 2d systems from infinite projected entangled-pair states via deep anomaly detection. Sci. Post Phys. 11, 025 (2021).
ADS  MathSciNet  Article  Google Scholar 
Kottmann, K., Metz, F., Fraxanet, J. & Baldelli, N. Variational quantum anomaly detection: Unsupervised mapping of phase diagrams on a physical quantum computer. Phys. Rev. Res. 3, 043184 (2021).
CAS  Article  Google Scholar 
Cross, A. W., Bishop, L. S., Smolin, J. A. & Gambetta, J. M. Open quantum assembly language, arXiv preprint arXiv:1707.03429 (2017).
Smith, R. S., Curtis, M. J. & Zeng, W. J. A practical quantum instruction set architecture, arXiv preprint arXiv:1608.03355 (2016).
Nielsen, M. A. & Chuang, I. L. Quantum Computation and Quantum Information (Cambridge University Press, 2000).
Bilkis, M. An implementation of VAns: A semi-agnostic ansatz with variable structure for quantum machine learning. https://github.com/matibilkis/qvans.
Nam, Y., Su, Y. & Maslov, D. Approximate quantum fourier transform with o (n log (n)) t gates. NPJ Quantum Inf. 6, 1–6 (2020).
Article  Google Scholar 
Gibbs, J. et al. Long-time simulations with high fidelity on quantum hardware, arXiv preprint arXiv:2102.04313 (2021).
Geller, M. R., Holmes, Z., Coles, P. J. & Sornborger, A. Experimental quantum learning of a spectral decomposition. Phys. Rev. Res. 3, 033200 (2021).
CAS  Article  Google Scholar 
Fletcher, A. S., Shor, P. W. & Win, M. Z. Channel-adapted quantum error correction for the amplitude damping channel. IEEE Trans. Inf. Theory 54, 5705–5718 (2008).
MathSciNet  MATH  Article  Google Scholar 
Fletcher, A. S., Shor, P. W. & Win, M. Z. Structured near-optimal channel-adapted quantum error correction. Phys. Rev. A 77, 012320 (2008).
ADS  Article  CAS  Google Scholar 
Kosut, R. L., Shabani, A. & Lidar, D. A. Robust quantum error correction via convex optimization. Phys. Rev. Lett. 100, 020502 (2008).
ADS  PubMed  Article  CAS  Google Scholar 
Kosut, R. L. & Lidar, D. A. Quantum error correction via convex optimization. Quantum Inf. Process. 8, 443–459 (2009).
MathSciNet  MATH  Article  Google Scholar 
Taghavi, S., Kosut, R. L. & Lidar, D. A. Channel-optimized quantum error correction. IEEE Trans. Inf. Theory 56, 1461–1473 (2010).
MathSciNet  MATH  Article  Google Scholar 
Lloyd, S. & Weedbrook, C. Quantum generative adversarial learning. Phys. Rev. Lett. 121, 040502 (2018).
ADS  MathSciNet  CAS  PubMed  Article  Google Scholar 
Dallaire-Demers, P. L. & Killoran, N. Quantum generative adversarial networks. Phys. Rev. A 98, 012324 (2018).
ADS  CAS  Article  Google Scholar 
Caro, M. C. & Datta, I. Pseudo-dimension of quantum circuits. Quantum Mach. Intell. 2, 14 (2020).
Article  Google Scholar 
Bu, K., Koh, D. E., Li, L., Luo, Q. & Zhang, Y. Statistical complexity of quantum circuits. Phys. Rev. A 105, 062431 (2022).
ADS  MathSciNet  CAS  Article  Google Scholar 
Bu, K., Koh, D. E., Li, L., Luo, Q. & Zhang, Y. Effects of quantum resources on the statistical complexity of quantum circuits, arXiv preprint arXiv:2102.03282 (2021).
Bu, K., Koh, D. E., Li, L., Luo, Q. & Zhang, Y. Rademacher complexity of noisy quantum circuits, arXiv preprint arXiv:2103.03139 (2021).
Gyurik, C., van Vreumingen, D. & Dunjko, V. Structural risk minimization for quantum linear classifiers, arXiv preprint arXiv:2105.05566 (2021).
Caro, M. C., Gil-Fuster, E., Meyer, J. J., Eisert, J. & Sweke, R. Encoding-dependent generalization bounds for parametrized quantum circuits. Quantum 5, 582 (2021).
Article  Google Scholar 
Chen, C. C. On the expressibility and overfitting of quantum circuit learning. ACM Trans. Quantum Comput. 2, 1–24 (2021).
MathSciNet  Article  Google Scholar 
Popescu, C. M. Learning bounds for quantum circuits in the agnostic setting. Quantum Inf. Process. 20, 1–24 (2021).
MathSciNet  Article  Google Scholar 
Cai, H., Ye, Q. & Deng, D. L. Sample complexity of learning parametric quantum circuits. Quantum Sci. Technol. 7, 025014 (2022).
ADS  Article  Google Scholar 
Mohri, M., Rostamizadeh, A. & Talwalkar, A. Foundations of Machine Learning (MIT Press, 2018).
Vershynin, R. High-Dimensional Probability: An Introduction with Applications in Data Science (Cambridge University Press, 2018).
White, S. R. Density matrix formulation for quantum renormalization groups. Phys. Rev. Lett. 69, 2863 (1992).
ADS  CAS  PubMed  Article  Google Scholar 
Orús, R. A practical introduction to tensor networks: Matrix product states and projected entangled pair states. Ann. Phys. 349, 117–158 (2014).
ADS  MathSciNet  MATH  Article  CAS  Google Scholar 
Vidal, G. Classical simulation of infinite-size quantum lattice systems in one spatial dimension. Phys. Rev. Lett. 98, 070201 (2007).
ADS  CAS  PubMed  Article  Google Scholar 
Ferris, A. J. & Vidal, G. Perfect sampling with unitary tensor networks. Phys. Rev. B 85, 165146 (2012).
ADS  Article  CAS  Google Scholar 
Spall, J. C. An overview of the simultaneous perturbation method for efficient optimization. Johns Hopkins Apl Technical Digest 19, 482–492 (1998).
Google Scholar 
Kübler, J. M., Arrasmith, A., Cincio, L. & Coles, P. J. An adaptive optimizer for measurement-frugal variational algorithms. Quantum 4, 263 (2020).
Article  Google Scholar 
Browne, D. E. Efficient classical simulation of the quantum fourier transform. New J. Phys. 9, 146 (2007).
ADS  Article  Google Scholar 
Download references
M.C.C. was supported by the TopMath Graduate Center of the TUM Graduate School at the Technical University of Munich, Germany, the TopMath Program at the Elite Network of Bavaria, and by a doctoral scholarship of the German Academic Scholarship Foundation (Studienstiftung des deutschen Volkes). H.H. is supported by the J. Yang & Family Foundation. M.C., K.S., and L.C. were initially supported by the LANL LDRD program under project number 20190065DR. M.C. was also supported by the Center for Nonlinear Studies at LANL. K.S. acknowledges support the Department of Defense. A.T.S. and P.J.C. acknowledge initial support from the LANL ASC Beyond Moore’s Law project. P.J.C. and L.C. were also supported by the U.S. Department of Energy (DOE), Office of Science, Office of Advanced Scientific Computing Research, under the Accelerated Research in Quantum Computing (ARQC) program. L.C. also acknowledges support from LANL LDRD program under project number 20200022DR. A.T.S. was additionally supported by the U.S. Department of Energy, Office of Science, National Quantum Information Science Research Centers, Quantum Science Center. This research used resources provided by the Los Alamos National Laboratory Institutional Computing Program, which is supported by the U.S. Department of Energy National Nuclear Security Administration under Contract No. 89233218CNA000001.
Open Access funding enabled and organized by Projekt DEAL.
Department of Mathematics, Technical University of Munich, Garching, Germany
Matthias C. Caro
Munich Center for Quantum Science and Technology (MCQST), Munich, Germany
Matthias C. Caro
Institute for Quantum Information and Matter, Caltech, Pasadena, CA, USA
Hsin-Yuan Huang
Department of Computing and Mathematical Sciences, Caltech, Pasadena, CA, USA
Hsin-Yuan Huang
Information Sciences, Los Alamos National Laboratory, Los Alamos, NM, 87545, USA
M. Cerezo & Andrew Sornborger
Center for Nonlinear Studies, Los Alamos National Laboratory, Los Alamos, NM, 87545, USA
M. Cerezo
Joint Center for Quantum Information and Computer Science, University of Maryland, College Park, MD, 20742, USA
Kunal Sharma
Quantum Science Center, Oak Ridge, TN, 37931, USA
Andrew Sornborger
Theoretical Division, Los Alamos National Laboratory, Los Alamos, NM, 87545, USA
Lukasz Cincio & Patrick J. Coles
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
You can also search for this author in PubMed Google Scholar
The project was conceived by M.C.C., H.-Y.H., and P.J.C. The manuscript was written by M.C.C., H.-Y.H., M.C., K.S., A.S., L.C., and P.J.C. Theoretical results were proved by M.C.C. and H.-Y.H. The applications were conceived by M.C.C., H.-Y.H., M.C., K.S., A.S., L.C., and P.J.C. Numerical implementations were performed by M.C. and L.C.
Correspondence to Matthias C. Caro.
The authors declare no competing interests.
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
Caro, M.C., Huang, HY., Cerezo, M. et al. Generalization in quantum machine learning from few training data. Nat Commun 13, 4919 (2022). https://doi.org/10.1038/s41467-022-32550-3
Download citation
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41467-022-32550-3
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
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

Box Office: ‘Dragon Ball Super: Super Hero’ Crushes Idris Elba’s ‘Beast’ With $21 Million Debut – Variety

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

By Rebecca Rubin
Film and Media Reporter
“Dragon Ball Super: Super Hero” topped the box office in its debut, beating expectations by collecting an impressive $21 million in North American ticket sales.
The anime film, playing on 3,007 screens, is being distributed domestically by Crunchyroll, which specializes in anime film and television. “Super Hero” is a necessary bright spot in an otherwise dreary August at the box office. The newest “Dragon Ball Super” installment earned twice as much as the weekend’s other new nationwide release, Universal’s survival thriller “Beast,” starring Idris Elba. Dinged by mixed reviews, “Beast” opened to a modest $11.5 million from 3,743 North American cinemas.
“We’re absolutely thrilled that ‘Dragon Ball’ fans could come together to experience and enjoy this amazing film in theaters,” Mitchel Berger, Crunchyroll’s senior VP of global commerce, said in a statement. “Crunchyroll thanks all of the fans, whether or not you are a ‘super’ fan or a newcomer, and we hope they come back again and again.”

“Dragon Ball Super: Super Hero” is the latest example of the passionate audience for anime films in the United States. Crunchyroll, which is owned by Sony Pictures, has been owning the market in North America. Earlier this year, the company’s PG-13 “Jujutsu Kaisen 0” notched a remarkable $17.6 million in its debut, while its 2021 release “Demon Slayer the Movie: Mugen Train” generated $21.2 million to start, which is even more impressive since cinemas were still operating at reduced capacity at the time. However, anime features tend to play like horror films in terms of ticket sales, with front-loaded performances begetting substantial declines after the first week of release.
Directed by Tetsuro Kodama, the well-reviewed “Dragon Ball Super: Super Hero” benefitted from its presence in premium formats, including Imax, 4DX and Dolby Cinemas. The movie played on 327 Imax screens, which accounted for $3.4 million in domestic ticket sales. At Imax, those returns rank as the widest and highest-grossing opening weekend for an anime film. For CJ 4DPlex, a South Korean company that specializes in immersive multi-sensory technology, “Dragon Ball Super: Super Hero” also set an anime record in the format with $525,000 from 51 domestic screens.
“This is another outstanding Crunchyroll anime opening. This has become an impressive niche theatrical business,” says David A. Gross, who runs the movie consulting firm Franchise Entertainment Research. But, he adds, “Crunchyroll movies play fast in the U.S.; their domestic multiples are low.”
Like critics, audiences were similarly mixed on “Beast,” giving the film a so-so “B” CinemaScore. Directed by Icelandic director Baltasar Kormákur (“Adrift”), “Beast” follows a father and his two teenage daughters (Iyana Halley and Leah Jeffries), who find themselves hunted by a massive rogue lion. The movie, which cost $36 million to produce, has raked in $10.2 million to date at the international box office.
Jim Orr, Universal’s president of domestic distribution, is encouraged that “Beast” appealed to a broad audience. Among opening weekend crowds, 34% were African American, 26% were Caucasian, 23% were Hispanic and 10% were Asian. Meanwhile 47% of ticket buyers were above the age of 35, skewing slightly older than expected.

“It’s great to see,” Orr says, referring to the audience breakdown. “It’s a very good start to what I’m convinced will be a great run.”
Elsewhere on domestic box office charts, Sony’s action-thriller “Bullet Train” slid to third place after two consecutive weekends at No. 1. The movie brought in $8 million from 3,781 locations in its third weekend of release, taking its domestic tally to $68.9 million. That’s a decent result for a star-driven action film outside of a franchise in today’s fractured moviegoing landscape. But it cost $90 million to make — and many millions more to promote to the masses — meaning it needs to keep chugging along in theaters to justify its hefty budget. “Bullet Train” has amassed $81 million overseas and a solid $150 million globally.
In fourth place, Paramount’s “Top Gun: Maverick” pulled in $5.85 million from 2,969 locations in its 13th weekend of release. Tom Cruise’s blockbuster action sequel, which hits home entertainment in the coming days, has grossed an eye-popping $683 million to date, enough to overtake Marvel’s 2018 superhero epic “Avengers: Infinity War” ($678 million) as the sixth-highest grossing domestic release in history. For those keeping track at home, it’s only $17 million away from catching the No. 5 spot, “Black Panther” ($700 million). Since it opened in theaters over Memorial Day weekend, “Maverick” has only spent one weekend outside of the top five on North American box office charts.
The Warner Bros. animated adventure “DC League of Super-Pets” took the No. 5 slot with $5.77 million from 3,537 venues. After four weeks in theaters, the family friendly film has generated $66.4 million in North America.
On the independent scene, A24’s slasher satire “Bodies Bodies Bodies” took 10th on domestic charts, adding $2.4 million from 2,541 theaters. The movie, which has been slowly expanding its footprint in recent weeks, has generated $7.4 million to date.
In far fewer theaters, Paramount’s psychological horror film “Orphan: First Kill” collected $1.6 million from just 498 venues. The prequel to 2009’s “Orphan” has such a scant screen count because it debuted simultaneously on digital platforms and on Paramount+, the studio’s streaming service.
A Variety and iHeartRadio Podcast
The Business of Entertainment

source

‘Joker’ & ‘Rebel Moon’ Sequels Among 18 Pics Set For California Tax Credits; Expected To Generate $915M In Production Spending – Deadline

By David Robb
Labor Editor
Four big-budget features are among the 18 film projects that have conditionally qualified for the latest round of California’s tax credit program. Together, they’re expected to generate an estimated $915 million in overall production spending across the state and employ nearly 600 actors, more than 2,500 crew members and thousands of background performers. See the list below.
“We are thrilled to continue welcoming the kind of big-budget films that used to be so susceptible to runaway production,” said California Film Commission executive director Colleen Bell. “In addition to our incentive, we have the best talent, crews, infrastructure, locations, weather, and so much more. California is ready to help filmmakers make the most of all we have to offer.”

Hollywood’s Unions Praise Gov. Newsom’s Pledge To Sign Bill Extending California’s Film & TV Tax Incentives Program

In addition to cast and crew, the 18 films are also expected to generate significant post-production jobs and revenue for California VFX artists, sound editors, sound mixers, musicians, and other workers and vendors.
The four big-budget films receiving the subsidies are:
Combined, the four films are expected to generate $377 million in qualified spending and $748 million in overall spending in California.

“No other city has more resources than L.A. when it comes to our industry, and to have this kind of support for the work we do from the California Film Commission’s Film & Television Tax Credit Program is a nice welcome for all of us on Joker: Folie à Deux,” said Todd Phillips, the film’s director/writer/producer. “Having spent the last several years filming in iconic locales here in the U.S. and around the globe, I’m looking forward to returning to work in the place I call home and to serving the state by bringing a project of this scope to California.”
The latest round of films selected to receive tax credits also includes 14 independent films – five with qualified spending of more than $10 million and nine with qualified spending of $10 million or less. None of those 14 films received more than $3 million in tax credits.
The commission has reserved $93.7 million in tax credit allocations for the 18 conditionally approved projects – more than $37 million of which is going to the two Netflix Production projects. The list of approved projects is subject to change, as applicants may withdraw from the tax credit program and their reservation of credits is reassigned to one or more projects on the wait list.
Production of the 18 projects is set to occur over a combined 718 filming days in California, with 218 days of planned shooting outside of Los Angeles. The Thomas Crown Affair plans to shoot 30 of its 54 filming days in San Francisco. The 1968 version of the film was shot in and around Boston, and the 1999 version was filmed in New York City.
The 18 films’ $915 million in overall production spending includes $503 million in “qualified” spending, which the Film Commission defines as wages to below-the-line workers and payments to in-state vendors. Only the qualified portion of each project’s budget is eligible for tax credits under the state’s incentive program.
The California Film Commission received a total of 57 applications during the July 18 – 25 feature film tax credit application period. The next application period for feature films will be held January 30 – February 6, 2023. The next application period for TV projects will be held September 19 – 26, 2022.
The current generation of the California Film and TV Tax Credit Program – dubbed “Program 3.0” – was launched on July 1, 2020. New provisions include a pilot skills training program to help individuals from underserved communities gain access to career opportunities. Program 3.0 also adds provisions requiring projects to have a written policy for addressing unlawful harassment and enhanced reporting of above and below-the-line cast and crew employment diversity data.
In 2014, the California legislature passed a bill that more than tripled the size of the state’s film and television production incentive, from $100 million to $330 million annually. Aimed at retaining and attracting production jobs and economic activity across the state, “Program 2.0” also extended eligibility to include a range of project types (big-budget feature films, TV pilots and one-hour TV series for any distribution outlet) that were excluded from the state’s first-generation tax credit program.
Program 2.0 also introduced a “jobs ratio” ranking system to select projects based on “qualified” spending (wages paid to below-the-line workers and payments made to in-state vendors). To spur production statewide, an additional 5% tax credit was made available to non-independent projects that shoot outside the Los Angeles 30-Mile Studio Zone or that have qualified expenditures for visual effects or music scoring/track recording. The five-year Program 2.0 went into effect on July 1, 2015 and wrapped its fifth and final fiscal year (2019/20) on June 30, 2020.
California Film Tax Credits May Generate $915M In Production Spending
Subscribe to Deadline Breaking News Alerts and keep your inbox happy.
Signup for Breaking News Alerts & Newsletters
Get our latest storiesin the feed of your favorite networks
We want to hear from you! Send us a tip using our annonymous form.
Sign up for our breaking news alerts
Deadline is a part of Penske Media Corporation. © 2022 Deadline Hollywood, LLC. All Rights Reserved.

source

In Football-Themed Ads, FanDuel Lays Odds on Just About Everything – Muse by Clio

“What is life but chance, and every moment in it a bet?”
With those words, FanDuel kicks off a cheeky campaign for football season that lays odds on just about everything in sight.
What are your chances of waking up on time? Enjoying the correct shower temperature? Finding a fresh carton of milk in the fridge? Or losing your cable-TV signal with the most crucial play of an NFL game about to unfold?
Heck, what was the likelihood of a big bang forming the cosmos and sparking the evolutionary chain leading to humanity’s existence in the first place?
A :90 from Wieden+Kennedy New York explores such questions (and many more) as “odds” flash on-screen. The message: “Sure things” don’t exist, and we take chances almost every hour of the day—so why not have some fun placing bets through FanDuel?
To view this video please enable JavaScript, and consider upgrading to a web browser that supports HTML5 video
“We want to appeal to all sports fans—those that are familiar with sports betting, and those newer to the category,” W+K creative director Gerard Caputo tells Muse. “The every-moment-is-a bet approach enables us to talk to everyone across this spectrum and have a point of view on everything from the creation of the universe to sports to popular culture. When the core idea is so far reaching and versatile, you can connect with almost anyone in meaningful ways.”
In a broader sense, the ads portray FanDuel as “a sports and entertainment brand” for consumers seeking deeper connections “beyond the offers and odds boosts commonly pushed in this category,” Caputo says.
So it’s a lifestyle play, with betting integrated into a day of watching football with friends. That’s an appealing approach and clearly on brand. (Though one could argue that since life’s so darn chancy, it might make sense to chill on the wagers and keep a few extra bucks in the bank!)
Director Nick Ball delivers a jaunty, well-paced narrative that really pops thanks to his distinctive use of first-person perspective. This technique “puts the audience in the point of view of the decision makers as they make all of these life bets throughout the campaign, so conceptually it fits our storytelling needs,” Caputo explains.
A half-dozen more ads will launch as the NFL season unfolds.
FanDuel archrival DraftKings embraces a glitzier approach, embedding its brand in original rap songs and casting celebrities like Martin Lawrence and Jerry Rice in a quest for mainstream buzz.
AGENCY : W+K NEW YORK
GLOBAL CHIEF OPERATING OFFICER: Neal Arthur
GLOBAL CHIEF CREATIVE OFFICER: Karl Lieberman
GROUP CREATIVE DIRECTOR: Gerard Caputo
COPYWRITER: Ray Ali
ART DIRECTOR: Jackie Anzaldi
HEAD OF DESIGN: Frank DeRose
DESIGNER: Mary Lamphier
HEAD OF INTEGRATED PRODUCTION: Nick Setounski
EXECUTIVE PRODUCER: Jess Griffeth
SENIOR PRODUCER: Jordan Leinen, Luke Rzewnick, Rob Marmor
PRODUCTION ASSISTANT: Stefan Puente                     
GROUP ACCOUNT DIRECTOR: Ryan Johnson
ACCOUNT DIRECTOR: Perry Kayla Fox     
ACCOUNT SUPERVISOR: Nico Rodiquez
ACCOUNT EXECUTIVE: Will Parker
SENIOR PROJECT MANAGER: Alex Doomany
GLOBAL GROUP STRATEGY DIRECTOR: Stéphane Missier
SENIOR STRATEGIST: David Ceng
COMMUNICATIONS PLANNING DIRECTOR: Zack Green
SENIOR COMMS STRATEGIST: Javaad Beg
SENIOR BUSINESS AFFAIRS MANAGER: Julie Petruzzo, Paul FitzGerald
BUSINESS AFFAIRS MANAGER: Joshua Christie
TRAFFIC SUPERVISOR: Sonia Bisono
TRAFFIC MANAGER: Laura Guerra
PRODUCTION COMPANY: MJZ
DIRECTOR: Nick Ball
EXECUTIVE PRODUCER: Emma Wilcockson
VIRTUAL MANAGING DIRECTOR/ PARTNER: Scott Horan
LINE PRODUCER: Line Postmyr
VIRTUAL LINE PRODUCER: Joan Garcia
PRODUCTION MANAGER: Kevin Dirinpour
BIDDER: Martha Lucas
DIRECTOR OF PHOTOGRAPHY: Roman Vasyanov
PRODUCTION DESIGNER: Olly Williams
LA STYLIST: Roger Forker
1st AD: Tina  Fratnik
CASTING DIRECTOR: Norma Masip
STORYBOARD ARTIST: Josh Hayes
EDITORIAL COMPANY: WORK EDITORIAL
Executive Producer: Alejandra Alarcon
Managing Director: Erica Thompson
PRODUCER: Samara Kelly
PRODUCER: Victor Medina
EDITOR: Jono Griffith
EDITOR: Dan De Winter
EDITOR: Trish Fuller
Sr. Cutting Assistant: Chris Wronka
Sr. Cutting Assistant: Audrey Weiner
St. Cutting Assistant: Chris O’Brien
ASST. EDITOR for Jono: Miles Watson
ASST. EDITOR for Dan: Julian Laing
ASST. EDITOR  for Trish: David Soto
VFX COMPANY: PARLIAMENT
LEAD  SUPERVISOR: Thomas Graham
SENIOR PRODUCER: Will Jellicorse
SENIOR PRODUCER: Jess Ambrose
Production Coordinator: Kate Matos
2D Lead: John McIntosh
COLORIST COMPANY: COLOR COLLECTIVE
Executive Producer: Claudia Guevara
Colorist: Alex Bickel
Colorist: Mike Howell
Color Assistant: Javiera Ergas
RECORD AND MIX: SONIC UNION 
STUDIO DIRECTOR: Justine Cortale
MIX ENGINEER: Fernando Ascani
MIX ENGINEER: Julienne Guffain
SOUND DESIGN & ORIGINAL MUSIC: Barking Owl
Creative Director: Kelly Bayett
Executive Producer: Ashley Benton
Sound Designer: Stuart Welch
Assistant Sound Designer: Mikayla Petersen
The best in creativity delivered to your inbox every morning.
Sign up for the daily Museletter for the latest ad campaigns and the stories behind them.
MUSE by Clio © 2022

source

Emma Ogosi Clocks 80 – THEWILL NEWS MEDIA


August 22, (THEWILL) – Quintessential music artiste and producer, Emma Ogosi, is 80 years. One would have expected him to roll out the drums in celebration, but he had a quiet day with family and friends. Good wishes from fans all over the world, who appreciate his music and still remember the good old days, poured in on social media as they celebrated the legacies of the great icon.
Ogosi was a former Air Force Cadet Officer for some years and performed with his band at his junior officer-friend, Ibrahim Badamasi Babangida’s wedding to Lady Maryam in 1969. Having paid his dues in the industry, he became one of the pioneers of music production with Midwest Television, now NTA Benin.
Although he studied Aircraft Engineering at the German Air Force Academy, his love for music made him venture into the entertainment industry and discovered one of the greatest talents, Evi-Edna Ogholi, who rocked the music scene with her reggae songs in the 1980s and 1990s. He was her producer for many years and the two eventually got married and welcomed two sons. However, their marriage hit the rock with Ogholi citing domestic violence as then reason for her marital woes. Ogholi later quit the music scene when the ovation was loudest and relocated to Paris, doing music and volunteer work with United Nations Children Fund (UNICEF) in Paris, France.

 

source

Masters changes tune on McConnell, says he's hopeful GOP leader will offer financial support in Arizona – Yahoo News

Masters changes tune on McConnell, says he’s hopeful GOP leader will offer financial support in Arizona  Yahoo News
source

Neumann bridges the gap with KH 150 – Worship AVL

Neumann bridges the gap with KH 150
The Neumann KH 150 is a bi-amplified DSP-controlled loudspeaker that joins the manufacturer’s line of reference-class studio monitors. Featuring the same resolution as all other Neuman Studio monitors, the speaker integrates a newly developed 6.5-inch woofer with what the manufacturer describes as “ultra-low distortion and high SPL capability”. Despite the cabinet’s diminutive size, the KH 150 has a flat frequency response from 39Hz to 21kHz (±3 dB) and is recommended for all critical listening applications, from broadcast to music production, including styles that require a deep low-end or elevated listening levels.
“The KH 150 closes the gap between the KH 120 and the KH 310, but it does far more than that,” explained Neumann portfolio manager Stephan Mauer. “Its internal DSP engine offers many benefits such as a phase linearised crossover and a new level in tonal consistency. The KH 150 is made to extremely low tolerances of no more than ±0.8dB. And you can hear it, too: Stereo imaging is razor-sharp.”
The digital electronics housed within the cabinet allows the KH 150 to be calibrated via Neumann’s MA 1 Automatic Monitor Alignment. In addition, the unit is said to be “extremely energy efficient”, with Neumann engineers having developed a patent-pending amplifier technology that combines superior audio performance with the energy efficiency of Class-D. Although the KH 150’s dual amplifiers deliver up to 145W to the 6.5-inch woofer and 100W to the 1-inch tweeter, they reportedly only consume 17W when idle. Auto-standby reduces power consumption to 0.3W when the KH 150 has not been used for a period of time.
Connectivity on the cabinet comprises of analogue and digital inputs (S/PDIF). The monitor will also be available in an AES67 variant with redundant AES67 audio network ports that are fully compliant with broadcast standards such as ST 2110, ST 2022-7 redundancy and RAVENNA. At the same time, the variant will be compatible with Dante-generated AES67 network streams.
You may also like:
By using our website, you agree to our use of cookies. Learn more
Cookies are short reports that are sent and stored on the hard drive of the user’s computer through your browser when it connects to a web. Cookies can be used to collect and store user data while connected to be able to provide you the requested services. Often cookies are getting deleted when the user leaves a site or logs out of it.
There are several types of cookies:
When you keep using our website, in compliance with Article 22 of Law 34/2002 of the Information Society Services, in the analytical cookies treatment, we have requested your consent to their use.
We use cookies to improve our services. For more details please refer to our Terms of Use and/or our Privacy Policy.
Please note that you can enable or disable and delete cookies in your web browser.

source

Explaining Hollywood: How to get a job as a set decorator – Los Angeles Times

The two-time Emmy-winning HBO series “Silicon Valley” follows a group of programmer friends working together to establish a company, Pied Piper.
In the last two seasons of the series, Brandi Kalish was hired as the show’s set decorator and tasked with giving character to the Pied Piper employee workspace.
“We had to sit at each desk and create a personality — like, this person is super into toys or LEGOs, this person is super organized or this desk only has a water bottle on it,” Kalish said.
To get some decorating inspiration for the hundreds of employee desks she and her team had to embellish for the show, Kalish went to San Francisco and toured several tech offices. As she visited these workspaces, she snapped reference photos to share with her team.
Kalish said it feels like putting puzzle pieces together on set.
A set decorator takes a script and creates a list of items — for example, furniture, lighting and artwork — for each set or environment that will be filmed. The decorator also will get direction and a budget from the projects’ showrunner, director and sometimes director of photography.
The Times interviewed Kalish and three fellow set decorators — Andrew Baseman (“Crazy Rich Asians” and “Severance”), Rae Deslich (“Promising Young Woman” and the reboot of “Supermarket Sweep”) and Rosemary Brandenburg (“Spider-Man: No Way Home” and “CastAway”) — to learn what it takes to become a set decorator and how to get into the industry.

Do you have big Hollywood dreams? This article is part of a series on starting and building entertainment industry careers. Read on.
More coverage

Inside the business of entertainment
The Wide Shot brings you news, analysis and insights on everything from streaming wars to production — and what it all means for the future.
You may occasionally receive promotional content from the Los Angeles Times.

Growing up in the Midwest, Kalish said she was “constantly coming up with ideas on how to redo my room.”
She went through a Southwest phase, which involved desert colors and asking her dad to sandblast cactuses on her windows — and then an all-black phase that included black window treatments and a black bedspread. Her love for textiles, furniture and art influenced her to go to art school for design.

For the record:
10:03 a.m. Aug. 17, 2022An earlier version of the article stated that Brandi Kalish dated a member of the band Korn. She dated a member of the band’s touring team. It also stated the band’s production designer needed a set decorator; it was the band’s manager.
In 1999, she was dating a member of Korn‘s touring team, and the band’s manager needed a set decorator to help transform the stage to look like a Gothic church.
Kalish looked in the phone book under “p” for prop house and went there with Korn’s credit card. She rented ornate candelabras, rugs, chandeliers and church kneelers.
All the pros The Times spoke to agree that a passion for art and the world around you is a necessity.
Deslich said you need a love for physical objects — and arts and crafts.
When they were a teenager, Deslich loved going to the renaissance fair but didn’t have a lot of money to buy costumes. They realized chainmail could be made with wire and a lot of patience.
“There’s a correlation between someone who looks at something and is like, ‘I think I can see how that’s made,’ and somebody who looks at a room and sees how a room is just a combination of all its elements,” Deslich said.
It’s about having an understanding that the world is movable, they said. You can pick up that piece of furniture and rearrange it or get it out of a space entirely if you don’t like it.
Set decorators also have to be organized, because whether you’re working on a film or a television series, you’ll be juggling multiple sets. Baseman has physical files of each of his sets. The files contain item lists, reference photos and other documents.
It’s also important to be curious and flexible.
Set decorating for “Crazy Rich Asians” took Baseman to Malaysia to learn about Peranakan culture. It also was an opportunity for him to work with an all-new local crew of buyers and assistants, instead of his go-to prop houses and shops.
“I really relied on them, and they found incredible pieces and showed me the lay of the land,” he said.
Company Town
Production accountants are in high demand in the entertainment industry and can pull in high salaries with the right experience. Plus, you don’t have to be a CPA.

Baseman’s parents were antique dealers, so he grew up with an appreciation for found objects and history. At Carnegie Mellon University, where he earned a bachelor’s degree in set and costume design, he started arranging and creating sets for theater, which led to opportunities on Broadway.
Many of the pros The Times talked to went to school — Kalish went to art school, and Deslich went to film school, thinking they wanted to be a director, writer or producer — which was helpful, but they say that there are many different paths, and it’s most important to fill whatever gaps you may have.
If you didn’t go to film school, read books on film theory, Deslich suggested.
Brandenburg majored in drama at Kenyon College in Ohio, but she said it was her real-world experience — exploring museums and historic sites in Europe as a kid when she traveled with her family on work trips — that influenced much of her early theater stage decoration.
A good resource for learning about set decorators is the Set Decorators Society of America. The guild provides resources, educational opportunities and networking opportunities with pros in the field.
Company Town
What does it take to dress characters in a film a or television show, and how do you get started if storytelling through clothing is your passion? Hollywood costume designers explain.

All set decorators that The Times spoke to said there are many different ways of getting into the industry, but common ones involve working as a production assistant, moving up to assistant set dresser (and then lead man, which is the lead dresser) or becoming a buyer. Either track can lead to becoming an assistant set decorator before getting a shot to lead the team.
Deslich has worked as a production designer but said they found their niche as a set decorator. “Mostly I just want to decorate bigger and weirder projects,” Deslich said.
Set decorators also can work in different mediums.
After working on tour with Korn, Kalish started as an intern for production designer Michael Whetstone (“Curb Your Enthusiasm” and “Abbot Elementary”) at Whetstone Designs. She started out decorating stages for Britney Spears, Ice Cube and Eminem. Working with the bands led to music videos, which led to commercials and then to TV and film.
Similar to Kalish, after Baseman worked on Broadway, he started getting calls to decorate commercial sets and then moved on to television and film sets.
Company Town
Production designers are both creatives and bosses.

It’s common to start out as a set decorator on nonunion productions with microbudgets, working for little or no pay, Deslich said.
Once you have more experience under your belt, you can start working on the requirements to join the Affiliated Property Craftspersons IATSE Local 44 union.
To join, Baseman said it’s ideal to have experience as a production assistant and take union-provided classes specific to set decorators. You’re also required to take a test.
He said once you make the jump into the union, you get pension and healthcare benefits and a union-negotiated pay rate.
Baseman acknowledged that it takes time to get into the union, but he encourages newcomers to keep at it.
“Once you’re in, you’re good, and you’ll continue to work,” he said.
Once you gain enough experience, you’ll be able to negotiate for higher pay than the rates set in the union contract, Brandenburg said.
Company Town
The first step in your Hollywood career shouldn’t be paying L.A. rent. Here’s how to take your first steps toward a career in the entertainment industry, according to experts.

High-definition television has changed the way set decorators design a scene, because viewers now can see every detail, Baseman said.
“I heard stories [about] soap operas when they were taped back in the ’60s and ’70s, where it was so low-resolution that rather than have rugs on the floor, they would just paint the floor to look like a rug,” he said.
It would save the project money. Now you can see “every fiber in the rug,” Baseman said.
Set decorators also have to keep lighting in mind when creating a specific scene or environment.
Brandenburg said because digital cameras are very sensitive to light, modern productions ramped up the use of the “practical lighting” technique. Rather than using a movie light overhead or off-camera, set decorators find desk lamps or hang fixtures to light the scene.
Set decorators work with the director of photography and the gaffer to make sure they provide the right light source, Baseman said. Does it need to be a hard lamp shade? Is an incandescent or fluorescent light better?
Smartphones also have changed the way set decorators can do research in the field. Kalish said she used to carry art books with her but now she can easily find reference photos online. She also can send photos to a showrunner for feedback.
Company Town
For many people who pursue entertainment as a career, it takes years to get yourself to where you are making money from your creative work. For making money in the meantime, there’s always waiting tables. But more and more people are turning to platforms like TikTok, Twitch and Patreon.

When young people start on set as a production assistant, Baseman said, there’s often a misconception that they’ll get to meet the actors and become friends with them. That’s not the case.
Baseman never met the stars of “Hocus Pocus 2,” a project he worked on last year.
When the cast arrived to film their scenes, the set was already decorated, and Baseman was already working on the next set.
“I’m always focused on the work, and I’m always two steps ahead of them,” he said.
Also, as newcomers start working on sets, they will hear “Just do anything” to get experience. But always try to work on something that is close to your goal, said Deslich.
When you are on a set, don’t get taken advantage of, they said. For example, don’t let a production work you for 20 hours in a row without feeding you or taking care of you.
Don’t be afraid to walk away, they said.
“There’s this perception of, ‘Oh, it’s a small town and you don’t want to burn bridges,’” Deslich said. “But no, it’s actually a very big town.”
Company Town
It takes a team to design the hair in a film or television show, whether the goal is to impress viewers with intricate styles or to make the hair so seamless that no one thinks about it.

Never stop researching: When she’s not arranging a set, Kalish said, she’s always “on a project with herself” — reading books, flipping through magazines or looking through Tumblr, Instagram and Pinterest for research.
Kalish will “go to buildings and travel to look at places I’ve never seen before or [revisit] places because it’s just a constant input that I’m collecting in my brain,” she said.
Use references from your own life experience: Baseman said he got a call for a project that was set in the 1980s. He was a young adult then and could recall certain items and pieces from those years for his decoration design.
Understand that your view of the world might not be the same as someone else’s: Brandenburg remembers going through a set walk-through of a middle-class home with a director. They had two different ideas of what middle class should look like, and she had to up-market the set to match what the director was looking for. You have to have an open mind when collaborating with others on the creation of a set, she said.
Company Town
Send the L.A. Times your questions about breaking into and working in the entertainment industry.

Inside the business of entertainment
The Wide Shot brings you news, analysis and insights on everything from streaming wars to production — and what it all means for the future.
You may occasionally receive promotional content from the Los Angeles Times.
Follow Us
Karen Garcia is a reporter on the Los Angeles Times’ Utility Journalism team, which publishes stories and information that help people solve problems, answer questions and make big decisions about life in and around Los Angeles.

Company Town

Company Town

Music

Company Town

Company Town

Podcasts

Company Town

Subscribe for unlimited access
Follow Us

source

Roth and Dubb Bring New Level of Luxury to Downtown Saratoga – Thoroughbred Daily News

TDN Live :
Click here to watch the Goffs UK Premier Yearling Sale
CLICK HERE TO SUBSCRIBE TO THE TDN FOR FREE!
Remember Ben with a gift to Thoroughbred aftercare
Subscribe for FREE to the Daily PDF or the News Alerts
Home » Archive » Shared News » Roth and Dubb Bring New Level of Luxury to Downtown Saratoga
Updated: August 19, 2022 at 10:17 am
The Adelphi Hotel in downtown Saratoga | Diana Pikulski
By
While racing draws thousands of visitors to Saratoga Springs every summer, its biggest appeal day in and day out, 12 months a year, is its graceful, bustling downtown. The city is rich in history, dating back to the 1770s and the Victorian style lends a charm that Saratoga not only strives to maintain, but often mandates through historic preservation codes.
Two buildings on Broadway in Saratoga Springs are now undergoing a huge transformation without losing their historic charm. The Adelphi Hotel, which recently underwent a major renovation, will be merged with the adjacent Rip Van Dam Hotel, expanding the number of luxury hotel rooms from 33 to 78 and adding 85 luxury apartments to the Rip Van Dam.
“We bought the Adelphi in about 2013,” said Larry Roth, a partner with his wife and daughter in LNJ Foxwoods Stable. “It was me and another group that is no longer involved with the Adelphi. The Adelphi at that time was a rundown hotel. It was owned by two individuals that graduated from Skidmore probably in the 1960s. They had the hotel only open two months a year. The rest of the other 10 months of the year, they would travel around the world finding antiques and bringing the antiques back to the hotel. So, when we bought the hotel in 2013, it was really run down, full of thousands of antiques, and we decided that we were going to run it for a year or two and then we were going to renovate it.”

Saratoga’s appeal and its offerings have expanded far beyond the track.
In a small geographic area of just 28 square miles, the City of Saratoga Springs boasts the oldest operating racecourse in the U.S., a harness track and casino, a 2,379-acre state park, four golf courses, a top concert and ballet and orchestra venue, a 6.178 square mile lake, historic mineral springs and baths, two colleges, and a 171-bed hospital.
In 1977, the Saratoga Springs Preservation Foundation was organized partly in reaction to the building of the shopping mall just north of the city limits. By protecting the historic facades and buildings downtown, Broadway and its adjoining streets have continued to maintain viable retail businesses and acclaimed restaurants and pubs when many upstate downtown districts have dried up.
The initial renovation of the Adelphi took years and two or three times what the group expected to pay, Roth said, but it re-opened in 2017 to great acclaim. But looming on the south side of the Adelphi was the Rip Van Dam, which had been through a tumultuous era of turnover and suffered from what seemed like a lack of vision and funding. It made sense to Roth that they purchase the Van Dam, which was in need of repair on the same scale. His partners did not share the vision.
“Their philosophy was not the same as mine,” said Roth. “I told them that I think we should buy the Rip Van Dam because the Rip Van Dam had a lot of property. Someone’s going to build something there, a hotel, an apartment building, and you don’t want to really compete with another hotel. They said to me, ‘You’re wrong, no one’s going to build there.’”
So, in 2019, Roth bought out his partners, became the sole owner of the Adelphi, and bought the Rip Van Dam.
“That’s when Michael [Dubb] got involved with me,” said Roth. “I knew Mike was in the horse business and I knew he was a great builder. So, I thought it’d be good to find out if he would be interested in working with me, being my partner on this project.”
“Larry and I had known each other for years,” said Dubb. “And we have a very close mutual friend. And, you know, when Larry correctly purchased the Rip Van Dam property to protect his investment at the Adelphi so he could control it, and that was a very smart move.
“It’s a major undertaking. And given my knowledge of Saratoga, my years at Saratoga, and my knowledge of construction, having built buildings like this before, although not 180-year-old buildings, Larry thought it would be a good idea for us to talk. And that’s how the partnership was born.”
“Our vision is to just have the most luxurious five-star hotel residence, restaurants, a destination point for people coming from either New York City, Boston, Montreal, Buffalo,” said Roth. “It’s all almost equal distance, about a three-hour ride from those locations. And just to have the number one destination in the Northeast.”
“For any city, a vibrant downtown is vital,” said former two-term Mayor of Saratoga Springs, Joanne Yepsen. “This city is successful because we care deeply about historic preservation and we have an appointed design review commission overseeing proposed development on top of a planning and zoning process.
“This project is a diamond in the rough,” said Dubb. “Saratoga is arguably the best town or small city in New York State. It’s a one-of-a-kind town, and this is a one-of-a-kind project right in the heart of town. The plan we had was to model this after The Plaza in New York City where you have a five-star hotel, but you also have residences. And the people who live in the residences can avail themselves and do avail themselves of all the hotel services. So here, there are two restaurants here which they would walk directly to through the building. There’ll be a gym. There’ll be lounges for the guests and the owners, a business center, and underground parking.”
The plan for the condominium apartments will be a mix of studio, one-bedroom, and two-bedroom luxury apartments appointed with hardwood floors, state-of-the-art kitchens, great appliances, marble bathrooms, and molding details. Each owner will have their own storage unit and a coveted parking spot underneath the building. The new hotel rooms will mirror the existing rooms at the Adelphi.
“The hotel units will be and are obviously very, very, very high-level luxury units,” said Dubb. “So, there’ll be heated bathroom floors, automatic blinds, suite-type living. Some of the new hotel rooms will have their own den and just all of the services and all the luxury that you would expect in a five-star hotel.”
In addition to the new hotel rooms and residences, there have been significant changes and upgrades to the common areas of the hotels, with more in the plan.
“We’ve increased the ballroom space. So, the capacity went from about 125 people to 225 people,” said Dubb. “That project is done. We’ve increased the size of Morrisey’s, put additional indoor dining rooms in, and also a sushi bar. And that project is done. The Salt and Char Restaurant, which is part of the building, is through phase one and the interiors of the existing Salt and Char has been completely rebuilt. But as construction progresses, there’ll be another dining room, about 800 square feet–state of the art. You know, with that real steakhouse feel, to go with the great food.
“In the original Adelphi, part of the lobby space was lost to the kitchen. The kitchen is moving away into the new structure. So, the lobby of the Adelphi now, which is beautiful, is going to double in size.”
While both Dubb and Roth are deeply ensconced in the upper echelons of the horse business, they’ve committed to the City of Saratoga and its growing year-round clientele with the Adelphi-Van Dam project. The underground parking garage for 100 cars, and major street and sidewalk and infrastructure improvements add to the entire downtown experience. And both partners are bullish on the long-term success of the project outside of the racing months.
“It’s not just the racing industry,” said Roth. “We’ve been extremely busy even during the off-season. Our occupancy during the off-season runs probably about 70%. During the track season, we’re running very close to 100%. Obviously, the July and August period is peak for us. But the rest of the year has, for the last year and a half, two years, has been truly outstanding.”
Preserving the historic façade and feel of the buildings seems to be more of a passion to the partners than a city requirement.
“The architecture and the character are beautiful. One great thing about Saratoga is its intense dedication to preserving its historic feel and its historic character,” said Dubb. “You really feel like you’re back in the early 1900s, in the late 1800s. And one thing we went to painstaking lengths to do is to preserve the entire facade of the existing Rip Van Dam Hotel. And previously, Larry did the same thing with the Adelphi Hotel. So, from the street, you know, I see turn-of-the-century pictures, not this century, the century before. And the facade will look just the same as it looked then.”
“The town is very pro-historical sites here,” said Roth. “We went before the board a number of times to show them what we’re doing. We made some revisions after they suggested we make some changes that they wanted us to do and we adhered to what they wanted us to do.”
What is the timeline?
“After we finish Salt and Char, we will be looking to get the next group of hotel rooms, some online for late fall this year, others on time for track season,” said Dubb.  “And that will be followed by the construction of the 85 condominiums. We’re looking at a time frame for completion of about 18 months. So, we’d be looking in its entirety at early ’24.”
The lure of Saratoga Springs, the two hotels so steeped in American history, and the racing industry at the Spa seem to give Roth and Dubb a deep sense of pride in the project. It was palpable as we walked around the busy establishment while they interacted with the staff and patrons. Individually and together, they appear relaxed and excited about what seems like a massive undertaking (and crater) in the middle of downtown.
“The combination, first the Adelphi, and then the Rip Van Dam and the adjoining land, is really a unique opportunity to marry two historic, legendary buildings together into one,” said Dubb. “A state-of-the-art five-star luxury hotel and residence building- nothing like this really exists in Saratoga and very few places like this exist anywhere in New York. So, this was really a unique opportunity. And when Larry reached out to me about it, my connection with Saratoga, you know, runs 50 years. And so, I’ve walked by these hotels for 50 years. I never dreamt I would have the opportunity to be involved and really make a difference in the City of Saratoga.”
“I don’t think I could have picked a better partner,” said Roth. “He is, first of all, a great friend. I think we have the same vision. We think alike. Being with him for the last two years on this project, I have learned so much from him about building, and I think even our contractor has learned a lot from Michael. He’s made a lot of suggestions to our contractor, which have panned out successfully.”
“I told Larry when we started, that it would be two very challenging years to build the building,” said Dubb. “COVID turned that into three and a half years. So, it’s challenging. I think we enjoy the challenge and I know when all is said and done, Larry and I are going to stand on Broadway, look up and be just the two proudest guys in Saratoga.”
Not a subscriber? Click here to sign up for the daily PDF or alerts.
This story was posted in Shared News and tagged , , , , , , , .


Subscribe for FREE to the Daily PDF or the News Alerts.
60 Broad St.
Suite 100
Red Bank, NJ 07701
p: (732) 747-8060
p: (800) 647-6571
f: (732) 747-8955
Editor / News Stories:
[email protected]
Advertising:
[email protected]
Customer Service:
[email protected]
Your TDN download has begun.
If the download does not complete, Click Here.
Never miss another story from the TDN
Click Here to sign up for a free subscription.

source

Pokémon GO Master and Senior Division Grand Champions crowned at Pokémon Worlds 2022 | Pokémon GO Hub – Pokémon GO Hub

Pokémon World Championships Event 2022 is still going strong, but the Pokémon GO portion of the event is now over, and for the first time ever, we got our Grand Champions in both the Senior and Master Divisions!
You can watch the entire Pokémon GO Worlds Finals here:


MEweedle fought his way through the Winners Finals, and faced off against Alan716 in the Senior Division’s Grand Finals. MEweedle fought hard, and Alan was no easy opponent, but in the event MEwedle was victorious! Curiously, MEweedle has opted not to use any Shadow Pokémon and two Steel types!
You can watch the entire battle set here. 
Do note that the Senior Division includes Trainers born in 2005, 2006, 2007, or 2008.

DancingRob faced off against a Korean Trainer named Godanhada in the Masters Division Grand Final, after winning the Losers Finals battle with CrescentAngels, who previously lost to Godanhada. 
Rob’s team included two Shadow Pokémon (Swampert and Venusaur), the new fan favorite Lickitung, and the usual Alolan Ninetales, Medicham and Galarian Stunfisk.
You can watch the entire battle set here. 
Do note that the Masters Division includes Trainers born in 2004 or earlier.
Pokémon GO Hub is the biggest Pokémon GO news site, publishing several informative guides, analysis, and news articles every month. Dedicated, focused and loving Pokémon GO.
© 2016 – 2022 Pokémon GO Hub | ALL RIGHTS RESERVED
Pokémon and Pokémon GO are copyright of The Pokémon Company, Niantic, Inc., and Nintendo. All trademarked images and names are property of their respective owners, and any such material is used on this site for educational purposes only. Solver d.o.o. has no affiliation with The Pokémon Company, Niantic, Inc., or Nintendo.

source

Alter Bridge announce 2023 North American tour – Next Mosh

ROCK AND METAL NEWS
Share the post “Alter Bridge announce 2023 North American tour”
Florida hard rockers Alter Bridge have announced a 2023 North American tour in support of their recently announced seventh studio album ‘Pawns & Kings’ (out October 14th via Napalm Records – pre-order). All dates are listed below.
Here’s more info from the trek announcement, “The 30-city tour – split into two legs – kicks off on January 25th in Tampa, FL and wraps up April 1st in Highland, CA, making stops in Nashville, TN (January 30th); Toronto, ON (February 11th); Kansas City, MO (March 11th) and Seattle, WA (March 23rd) to name a few markets. It has been almost three years since the quartet comprised of Myles Kennedy on vocals/guitars, Mark Tremonti on guitars/vocals, Brian Marshall on bass and Scott Phillips on drums has toured North America.”
They add, “Alter Bridge is also pleased to announce that Mammoth WVH – the Wolfgang Van Halen fronted band – will be joining them as direct support on the entire tour. Rockers Red will be playing on the first leg of the tour and newcomers Pistols At Dawn will be playing on the second leg. Tickets will be available everywhere starting Friday, August 26th.”
Tour Dates:
Jan 25 – TAMPA, FL – Seminole Hard Rock * [TIX]
Jan 27 – ORLANDO, FL – Hard Rock Live * [TIX]
Jan 28 – ATLANTA, GA – Coca Cola Roxy * [TIX]
Jan 30 – NASHVILLE, TN – Ryman Auditorium * [TIX]
Feb 1 – MONTCLAIR, NJ – The Wellmont Theater * [TIX]
Feb 2 – HUNTINGTON, NY – The Paramount * [TIX]
Feb 4 – SILVER SPRING, MD – The Fillmore Silver Spring * [TIX]
Feb 5 – PHILADELPHIA, PA – Franklin Music Hall *
Feb 7 – WALLINGFORD, CT – The Dome at Toyota Oakdale Theatre [TIX]
Feb 8 – BOSTON, MA – Roadrunner *
Feb 10 – WATERLOO, NY – Del Lago Casino Resort (On sale soon) ^
Feb 11 – TORONTO, ON – HISTORY ^ [TIX]
Feb 14 – DETROIT, MI – The Fillmore * [TIX]
Feb 15 – INDIANAPOLIS, IN – Murat Theatre at Old National Centre * [TIX]
Feb 17 – PRIOR LAKE, MN – Mystic Lake Casino Showroom * [TIX]
Feb 18 – CHICAGO, IL – The Riviera Theatre *
Mar 10 – QUAPAW, OK – Downstream Casino Resort #
Mar 11 – KANSAS CITY, MO – Arvest Bank Theatre at The Midland #
Mar 14 – CHESTERFIELD, MO – The Factory # [TIX]
Mar 15 – OKLAHOMA CITY, OK – The Criterion #
Mar 17 – HOUSTON, TX – 713 Music Hall # [TIX]
Mar 18 – DALLAS, TX – South Side Ballroom # [TIX]
Mar 20 – DENVER, CO – Mission Ballroom #
Mar 21 – SALT LAKE CITY, UT – Union Event Center # [TIX]
Mar 23 – SEATTLE, WA – The Paramount Theatre # [TIX]
Mar 25 – AIRWAY HEIGHTS, WA – Northern Quest Resort & Casino ^
Mar 28 – ANAHEIM, CA – House Of Blues # [TIX]
Mar 29 – TEMPE, AZ – Marquee Theatre # [TIX]
Mar 31 – RENO, NV – Silver Legacy’s Grande Expo Hall # [TIX]
Apr 1 – HIGHLAND, CA – Yaamava Theater at Yaamava Casino Resort ^
_____
^ Alter Bridge – Mammoth WVH
* Alter Bridge – Mammoth WVH – Red
# Alter Bridge – Mammoth WVH – Pistols At Dawn
Alter Bridge tour for 2023

‘Pawns & Kings’ track listing:
1) This Is War
2) Dead Among The Living
3) Silver Tongue
4) Sin After Sin
5) Stay
6) Holiday
7) Fable Of The Silent Son
8) Season Of Promise
9) Last Man Standing
10) Pawns & Kings
More Alter Bridge tour dates:
Nov 1 – HAMBURG, GERMANY – Sporthalle
Nov 2 – COPENHAGEN, DENMARK – Falconer [TIX]
Nov 4 – PARTILLE, SWEDEN – Partille Arena [TIX]
Nov 5 – OSLO, NORWAY – Sentrum Scene [TIX]
Nov 7 – KATOWICE, POLAND – MCK
Nov 9 – ESCH-SUR-ALZETTE, LUXEMBOURG – Rockhal [TIX]
Nov 11 – BERLIN, GERMANY – Columbiahalle
Nov 12 – HLAVNÍ MĚSTO PRAHA, CZECH REPUBLIC – O2 Universum [TIX]
Nov 14 – BUDAPEST, HUNGARY – Budapest Arena
Nov 16 – PARIS, FRANCE – Palais Des Sports [TIX]
Nov 18 – MADRID, SPAIN – Vistalegre
Nov 20 – BARCELONA, SPAIN – Razzmataz1
Nov 22 – MUNICH, GERMANY – Zenith
Nov 23 – ZÜRICH, SWITZERLAND – Samsung Hall
Nov 25 – ASSAGO, ITALY – Mediolanum Forum
Nov 26 – ZAGREB, CROATIA – Dom Sportova
Nov 28 – VIENNA, AUSTRIA – Wiener Stadthalle
Nov 30 – KÖLN, GERMANY – Palladium
Dec 1 – AMSTERDAM, NETHERLANDS – Ziggo Dome [TIX]
Dec 5 – NOTTINGHAM, ENGLAND – Motorpoint Arena [TIX]
Dec 6 – GLASGOW, SCOTLAND – OVO Hydro Arena [TIX]
Dec 8 – DUBLIN, IRELAND – 3 Arena [TIX]
Dec 9 – MANCHESTER, ENGLAND – AO Arena [TIX]
Dec 11 – BIRMINGHAM, ENGLAND – Resorts World Arena [TIX]
Dec 12 – LONDON, ENGLAND – O2 Arena [TIX]

Stream new video single “Silver Tongue”:

Artist photo by Chuck Brueckmann
Also Check Out: Vulnificus announce 2022 U.S. tour with Deterioration




Follow NextMosh On: Facebook, Twitter, Instagram, YouTube
This post may contain referral links.
Your email address will not be published.


















Delivered by FeedBurner



source

Subduxtion Tells Us More About the Behind the Scenes of His Remix of 'That Eskimo's Coat' – One EDM

Subduxtion continues surprising his followers with new hard-hitting tracks and impeccable mixes for his flagship weekly radio show ‘Dark Signals’. His fantastic and signature style and sound is displayed in his latest releases; most recently, subduxtion joined a group of talented DJs to remix Sizz Nickhrome’s track ‘That Eskimo’s Coat’. Subduxtion’s version undoubtedly showcases his sound and colors, an energetic beat, deep basslines and a mysterious but powerful overall vibe are the key components of this certified banger. 
While subduxtion continues to work on upcoming releases and episodes of his show, we asked him more about the behind-the-scenes process of his remix of ‘That Eskimo’s Coat’. 
Hi subduxtion, how are you? 
I’m good, thank you! It’s been a good year so far.  
Your remix of Sizz Nickhrome’s ‘That Eskimo’s Coat’ is out now, how did the collaboration come about? 
I’ve been a long-time fan and supporter of the label, Pocket Food Audio that released the remix collection. We first came to each other’s attention back in 2020. I reached out to them to do a podcast/mix for them in 2020. My podcast/mix was number 004. From there we just stayed in touch, which led to me contributing the track, ‘Dust’ to a compilation they released in late 2021. In 2022 they put out a call for remixers for ‘That Eskimo’s Coat’ and I definitely wanted to be part of the release. 
What elements of the original version inspired you to produce your remix? 
I used sections from the synth parts and the percussion. I used Serato Sample to make a number of two and four bar loops from the original elements and built my remix from those loops. I also sped-up the track and did some pitch shifting to change the key of the track. 
Did you experiment with new sounds or processes? 
The newest process that I’m using is utilizing guitar pedals as part of my sound-design process. Over the years I’ve collected a number of guitar pedals. I have a re-amp device which allows me to run hardware synths, virtual synths and drum machines through my various pedals. My pedal collection includes distortions, overdrives, delays, amp simulators, reverbs and a few pedals that can really mangle and manipulate audio in really interesting ways. 
What plugins did you use while producing this remix? 
I used Arturia Analog Lab 5, Air Hybrid 3, Native Instruments Maschine 3 and Battery 4, Valhalla VintageVerb,  Minimal Audio Rift, Cableguys ShaperBox, Waves HDelay as well several of the UA plugins – the 1176LN and SE, the Neve 1073 EQ and channel strip, the Pultec EQ Collection, the Studer A800 Multichannel Tape Recorder and the LA 3A and 2A plugins. In addition to the aforementioned plugins, hardware really played an important part of the remix. I made use of a Behringer DeepMind 12, an Arturia MicroFreak and a Roland TR-6S. I also used the JHS Pedals Colourbox 2, Game Changer Audio Plasma Drive (Eurorack version) and Light Pedal, and the TC Electronics June-60. The June-60 is a fantastic emulation of the much-loved Roland Juno-60 chorus. 
What would you like to achieve with this track? 
With every track that I create, the ultimate goal is to hear myself getting better as a producer. The thing we’re all chasing as artists is how close can I get what we hear in our head to be how the track finally sounds. 
Are there any other artists you’d like to collaborate with in the near future? 
I would love to collaborate with Bicep. I’m a huge, huge fan of theirs and it would be an incredible learning experience to go in the studio with them. The sound of their productions is like a masterclass in how to create space for a lot of elements to coexist in a mix. 
What keeps you inspired to make new music? 
I’m always listening to music whether new or old. Hearing what other producers and artists are doing is always inspirational. There’s always something I hear that makes me sit-up and take notice, whether it’s a particular sound, how the song is mixed or the arrangement, there is no shortage of inspiration. 
How would you describe your style and sound to someone who hasn’t listened to your music before? 
I would describe it as dark, moody Progressive House. I tend to favor minor chords and that instantly lends my tracks a darker and moodier sound. The sound of Progressive House from say, 1990 – 2010 certainly left its mark on me.  
What are you currently working on? 
Right now, I’m really focused on my weekly radio show ‘Dark Signals’. The show has been airing for seven months on a number of stations, worldwide. Beyond the radio show I’m taking a month or so to just experiment in the studio. I think it’s important to just get in the studio and play around without trying to write a new track. Just messing about can help clear your musical mind and give you some new found inspiration to start making new tracks.   
Listen and buy ’That Eskimo’s Coat (subduxtion Remix)’ now: 
https://www.beatport.com/track/that-eskimos-coat/16753621
Subduxtion Online: 
Instagram | Audius | Soundcloud | Beatport 
Sizz Nickhrome Online: 
Instagram | Soundcloud | Spotify 
Pocket Food Audio Online: 
Soundcloud | Beatport | Instagram 
Lorem ipsum dolor sit amet, consectetur.

source

Photos: LA startup is using recycled plastic to 3D print prefab homes – Business Insider

Source: Azure
Source: Insider, Insider 
Source: Azure
Source: Insider
Source: Azure
Source: Spectrum News 1 
Keep reading

source

WHAT IS FREEFORM BASS MUSIC? MERSIV BREAKS SHARES ALL YOU NEED TO KNOW – Magnetic Magazine

Peter Wallace
Freeform bass is a new genre of bass music that is taking over the Americas. From an outsider looking in, it can be hard to define precisely what this buddy blend of genres exactly is. But for an artist, as entrenched in the scene as Mersiv, what makes Freeform Bass so special is clear as day. 
So we decided that there is never a better time than now to bring the artist on to explain everything that you've ever wondered about Freeform Bass and answer all of the most popular questions I know you are asking about America's hottest genre dance music. 
Mersiv is a famous electronic touring artist known for Pretty Dark Loud, the title of his debut album released in 2021. Since he's toured nationally with Liquid Stranger and is preparing his cross-country run on The Mersiv Experience Tour. Grab tickets here.
Not From The Editor: The following is an Op-Ed from Anderson Gallegos, better known as Mersiv, on the origins of the Freeform Bass genre and its existence today.
The roots of Bass music originated from Jamaican Soundsystem culture due to artists like Lee "Scratch" Perry, later influenced by the Grime and Drum & Bass scenes in the UK. Freeform Bass expanded in the Electronic scene after Bass music had grown widely appreciated in the States post-2010.
Freeform Bass emerged in the U.S. after DnB, Dubstep, Trap, and other veins of Electronic music continued to find a home and grow here. After all, these genres had grown to be embraced more through live events, particularly at music festivals. 
Regional influences of Hip-Hop, Rock, and other live music began to evolve Bass music into a melting pot of styles.
There's a lot at play when discussing what's helped create community, both sonically and socially.
With its inherent acceptance and involvement of LGBTQ communities and many different cultures, the growth of rave and festival communities allowed for a more inclusive and welcoming environment for people to create. The utilization of modern technology to create immersive audio/visual experiences has promoted these styles of electronic music even more.
Liquid Stranger is a pivotal figure as the founder of WAKAAN, which has served as a label and platform for Freeform artists to create. My label, MorFlo Records, was created in a similar vein, empowering creatives to find their flow and their unique way of expression through sound.
Other influential figures are artists like PEEKABOO, Break Science, Pretty Lights, GRiZ, Manic Focus, and KOAN Sound.
Trey Sarmento
The "Pretty Dark Loud" idea was built over several years, but part of the actual record was written during the COVID lockdown. This gave me a rare break from shows to truly explore how I wanted to express myself.
Together, a wide range of time and emotions in life reflects in the art. It's about letting loose, not putting barriers in the music, and pulling from all sorts of inspirations.
Pretty Dark Loud communicated to people in meaningful ways through a wide range of sounds, vibes, emotions, influences, and tempos that didn't put into a box. Its ability to connect with people and pull from numerous inspirations places it under the banner of Freeform.
As mentioned, Freeform is an open range of tempos taking inspiration from multiple genres. Much of the music is blendable with other bass genres and fits well for hip-hop acapella use when mixing it live.
This usually lends to bpm signatures in 65-75 (130-150) bpm, 80-90 (160-180) bpm, and even sometimes glitch/mid-tempo/and house tempos in the 100-130 bpm range.
Trey Sarmento
Like my album Pretty Dark Loud, Freeform Bass music ranges from very melodic music to the more hard-hitting, chaotic sounds one might think of when they hear the term dubstep.
Freeform bass's versatility of styles comes from a broad spectrum of human emotion expressed through it and resonating with it. The space between the polarities of pretty and dark, subtle and loud, are explored throughout the genre.
Freeform music takes inspiration from all that came before while also exploring avenues for consciousness to manifest and reverberate as we explore the growing pains of moving into a new era of higher global consciousness.
Because of the nature of Freeform Bass, those looking to produce music in this genre should rely on a wide range of plugins, tools, and instruments. Freeform Bass has a diverse sound palette, and pulling synthetic sounds and pairing them with acoustic tones and found percussions can yield some exciting results. 
Synths like Massive, Serum, Phaseplant, and Omnisphere are necessities when shaping the sounds of Freeform Bass. Digital synths like these allow for precise automation moves, highly customizable sounds, and many other features that allow you to dial in the exact sound and vibe you are after.
Much like the VSTs mentioned above, Freeform Bass production can be taken to the next level when you pair synthetic drums' digital and hyper-sound-designed hits with the dusty, warm hits of live percussions. 
Layering the two styles of recordings and samples allows you to get the best of both worlds and helps arrive at a unique sound; giving you a better chance of having your music get noticed as the genre becomes more popular (and thus oversaturated)
Juxtaposed against the digital precision of VSTs like Serum or Omnisphere, using live instruments in your Freeform bass tracks can elevate your productions to new heights.
Recorded instruments have a unique texture, sample timings, and numerous other qualities that VSTs simply can't replicate. So whenever you have the option of having instruments on the track, take advantage of everything they offer. 
Whether you're using a clean and dialed-in bass line or a warm and distorted bassline, your low-end needs to be knocking, after all, it's called Freeform Bass! 
But unlike genres such as dubstep, the entire focus of Freeform bass doesn't necessarily have to be on the bass. This gives producers tons of creative freedom in their sound design, good selection, and mixing! 
Make sure the bass is enormous, and you'll be on the right track.
Producing Freeform Bass music, in my opinion, requires a firm foundation in music theory, sound design, mixing, and arrangement. Once you understand these principles, you must try new things and break your boundaries to tap into your flow-state to allow universal energy, thought, and emotion to manifest through your art.
At its basis, you're creating music that people can dance or flow to, so look for an auditory flow, rhythm, and "push and pull" effect within the journey of the song.
Many people knock on YouTube tutorials and say there is too much false information online when it comes to music production. 
You can learn anything you need to produce top-level Freeform Bass on YouTube, and I think it's one of the best things for the music-production community. 
Find inspiration in live events and the world around you. Remember that Freeform Bass is meant to be danced to, and the best way to see and hear what works in a live setting is to get out into the wild and experience it yourself. 
Utilizing EQ-ing, filtering, and automation to enhance the flow of the track. This is the best way to make your music move and maintain interest over the length of a song. 
It's the secret sauce to turning a song that sounds super 'loopy' into a living, breathing piece of music. 
Resampling is a powerful weapon that most producers don't start experimenting with until later in the game. 
Being able to design a sound, flip it, rework it, and reprint the audio opens up new levels of creativity that standard MIDI instruments can't achieve. 
Manipulating vocals and synths through modulation, especially with effects like reverb and delay, can be an excellent way to enhance the emotion and space of the track.
A track that feels empty and boring might just need a couple of layers of super-dialed-in reverb to feel full and punchy suddenly.
The difference between an amateur-sounding track that no DJ plays and a fantastic track on everyone's USB are all in the details. 
Utilize subtle amounts of foley and noise-type sounds to fill out each sound and the entire 3D frequency range. Pay extra attention to exactly how your automation is moving. Do a few more takes on your live instrument, so you have a ton of raw audio to resample and manipulate in the session. 
Attention to detail will help you out astronomically in the long run. 
Find a talented Mastering engineer to help you make your music presentable in the best way on extensive sound systems or any sound system.
Outsourcing this last step of the production process allows you to get an unbiased opinion on your final master and helps ensure that you get an objectively fantastic end result. 
Managing Editor at Magnetic Magazine – Former Marketing Manager at the Hyperbits Masterclass – Writer, Content Strategist, Music Producer, and Record Label Manager 

source

Andrey Sirotkin: "Air raid sirens have become a regular part of the daily routine" – Electronic Groove

EG caught up with Andrey Sirotkin to learn more about his current situation following the Russian invasion of Ukraine, the release of his ‘My Shelter / Nothing Is Safe’ EP, future plans, and much more.
Electronic Groove: Hi Andrey! Welcome to EG! How’s your year been so far?
Andrey Sirotkin: Hi! Thank you. Ever since Russia invaded Ukraine it’s been hard to keep my head above the water, but I’m as good as I can be in this situation.
Electronic Groove: Sorry to hear about this. A terrible situation that’s for sure. How do you keep up producing music while all this is going on?
Andrey Sirotkin: Producing music and rare local events where I take part is the only thing that keeps moving me forward. The rest is pretty demoralizing.

Electronic Groove: What do you think artists should do about political issues? Should they just keep focused on producing and staying creative or should they be vocal about what’s going on?
Andrey Sirotkin: First of all – artists are people, and they have their own issues and reaction to what’s happening. So to produce or to pause is their own choice and will. There are many ways in which we can spread information about the terrible things that Russian soldiers do in Ukraine and to Ukrainian civilians and children. Another thing that artists can do is charity/fundraising gigs and raise money by themselves. Of course, donating money to the Ukrainian arm forces is another way. So there’s room for all of those tools that artists can use. Everyone should participate in the way that works for them.
Electronic Groove: How did you get involved with dance music? And into music production?
Andrey Sirotkin: I have been listening to music since I was a little kid. During my school years, I bought a computer and started to edit music and then use music production software (more like games of course) to make my own music collages. By the end of my school years, I had a couple of tracks produced, but of course, they weren’t good enough to be released… In 2004 I took classes at the ‘Back 2 School’ DJ academy in Kyiv and ever since, I have been fully immersed in dance music.
Electronic Groove: Your latest release was ‘My Shelter / Nothing Is Safe’ EP. What can you tell us about this one in terms of production, equipment, and inspiration?
Andrey Sirotkin: This EP was produced while I hid with my family in a bomb vault in Lviv back in March/April. Both tracks are produced with a low-performance laptop and a pair of AirPods. That was the way to break through the feelings I had at that time.
Electronic Groove: Do you have any other releases for the rest of the summer season?
Andrey Sirotkin: Yep. These are: ‘Honest With Yourself / Love Is Forever’ EP with leftfield house and breakbeat. And, ‘Enter Reality 2.0’ EP with some deep techno, deep house, and a remix of a traditional Ukrainian lullaby produced by Zymosis. Almost all the productions I make with my synth from my studio, which I miss so much as I have left Kyiv. There’s also the release of my ‘Dusk’ EP on Default Position this September 12th. Four tracks of deep techno/electronic music. All proceeds from this EP will be donated to Red Cross Ukraine. Pre-order is now available.  All of these releases are my sublimation of the Russian-Ukrainian war and all that grief that Russians have caused.

Electronic Groove: How is the situation right now in your hometown? Are there any events going on? What’s day-to-day life like now?
Andrey Sirotkin: Air raid sirens have become a regular part of the daily routine. Some people stop to hide, and others, even while it’s on, continue to do whatever they were doing. Even though we’re not in the areas where the combats are at this moment, still we’re day-to-day in this informational field from the news of civilian people or children being injured or killed. And we all get news from our people who serve the country, or the news about them if they are injured or killed. So even being far from the hottest areas in conditionally safe places, we’re still affected by war and will be affected long-long time after it ends. The only events that take place now in Kyiv are the daily ones cause night is closed time by the government. All events gather money for the Ukrainian army (100% or some part from income). Mostly these are non-dance events, but slowly they’re getting back to the public.
Electronic Groove: Do you have any gigs planned for the coming months?
Andrey Sirotkin: In Ukraine, these are mostly spontaneous for me. Like, “can you play at our fundraising event please?”. And I’m ready to go and to do my best. Nothing is clear and can change very fast. On the other hand, I started 2 monthly radio shows at Gasoline Radio. ‘Against The Grain’ is for deep and non-mainstream electronic music, from deep techno and leftfield house/techno to ambient and downtempo. ‘Deviant Drums’ is a show about non-mainstream drum and bass and the diversity of its subgenres.
Electronic Groove: Moving into a more personal angle of your craft. How would you describe your sound by now? How was it evolved since you first started producing? What would you say are your main strengths and what would you like to improve?
Andrey Sirotkin: I came from drum and bass music and from 2004 made my way through dubstep to techno in 2013. After that, my sound evolved to the latest alias, (my real name) Andrey Sirotkin, which I use for all the music I make. Now the main focus of my music is deep techno and leftfield house and techno. Of course, some deviations will be included in future EPs. I’m falling into the black hole of modular synthesizers. That stuff really inspires me and drives my production the most. Of course, I produce some tracks 100% in the box by using samples only, but I always try something new for the next track, whether it’s a new synth, module, field recording, or a combination. My inspiration is to find something new for me, not to copy others’ sounds. And it’s the biggest problem, as most times I see it’s not fitting the classification by listeners and playlist curators 🙂
Electronic Groove: What are your favorite 3 producers and why?
Andrey Sirotkin: Robert Babicz / Rob Acid – a great artist with perfect taste and super skilled technically producer. Robert plays only his music and for me, he’s a true artist and a real human being! He inspires me a lot! The second would be Trevino – a techno project from Marcus Intalex (drum and bass producer). His music is not just some dry-sounding techno, but it’s breathing and constantly changing complex sound with non-cliché patterns and ideas. The third one is Paul Mac – another great producer whose music I deeply love. His musical ideas are awesome and his sound is so great that I love 9 of 10 of his tracks. Again, it’s not a dry cliche sound but moving against the grain.
Electronic Groove: Thank you for your time, Andrey! We wish you all the best for the future!
Andrey Sirotkin: Thank you for this interview, for your support of my releases, and for your support of Ukraine! Slava Ukraini!

Andrey Sirotkin’s new ‘Dusk’ EP will be out on Default Position. Purchase your copy here.
Follow Andrey Sirotkin: Soundcloud | Facebook | Instagram | Spotify | Patreon
Lucky Sun (aka Tom Lown) is best known for his super-deep productions which have been…
Tasha Blank is a celebrated international DJ, motivational speaker, and founder of BODY LVNGUAGE (formerly…
John Min was born and raised in Orestiada, Greece. Starting as a raver soon learned…

source

PIC micro-controller based synchronization of two fractional order jerk systems | 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: 14281 (2022)
Metrics details
The paper studies a 3D Chaotic Jerk oscillator with fractional derivatives. An approach is proposed to implement it on a PIC16F877A microcontroller in order to reduce the requirements for multiple analogue electronic components such as resistors, capacitors, coils, multipliers, operational amplifiers, which are very bulky and consume a lot of power. The behaviours of the underlying system are analysed analytically, numerically and experimentally. It comes from this analysis that the fractional model exhibits chaotic dynamics when for parameters for which the equivalent integer derivative system exhibits limit-cycles. The synchronization under two closed initial conditions is also studied, highlighting one of the most common applications of the chaos concept.
Concept of fractional calculus has become undoubtdly a subjet of active research field in nonliear science, due to its potential applications in electronics1, mechanics2, nuclear physics3, medicine4, financial systems5. Likewise, chaos has been also widely reported over the past century years6,7,8,9,10,11,12,13. It is well-known that chaotic systems have a very high degree of sensitivity to initial conditions and their evolution through phase space that appears unpredictable.
In fact, in 1963, Lorenz emphasized that a chaotic system solved with two very close initial conditions could have two completely different dynamics14. More recently, some electronic circuits exhibiting chaotic behavior are proposed in literature15,16,17. Furthermore, those implementing fractional order circuit have been also reported. For example, the fractional Chen circuit18,19, the fractional Chua circuit20 and the fractional Rössler circuit21. For a better review one can read22,23,24.
In truth, fractional dynamical systems are developed with the main idea of introducing the memory effect in the dynamics of the system. It is then observed that these systems present hidden attractors that the conventional approach does not exhibit25. Considering this particular advantage, it is therefore imperative to design fractional electronic circuits able to reproduce the desired behavior. To do this, an ordinary capacitor is replaced by a fractional capacitor whose impedance value must be determined. In this case, the Laplace transform of the differential operator (frac{1}{s}) is replaced by (frac{1}{s^m}), where (0<m<1) is the derivative order26,27,28,29,30,31.
However, to the best of our knowledge, it is arduous to determine the exact value of a fractional capacitor which corresponds exactly to his fractional derivative operator. Several techniques have been suggested to address this problem such as the Regular Newton’s Process32 and the Halley’s Iterative Method33. These methods were used to design the fractional circuits mentioned above using analogue component. From these circuits it can be seen that the analogue component approach requires very large capacitors and resistors which are difficult to find on the market. Furthermore, using large capacitors and large resistors means operating at high frequency34. We propose a numerical approach for the real implementation of a fractional system on a micro-controller based on Euler’s resolution method and applied to the synchronization of a Jerk system, which has never been done to our knowledge.
Jerk systems have important considerations for many applications in science and mechanical engineering. In35, it is noted that Jerk systems could exhibit several physical phenomena such as multi-stability, chaos or hyperchaos. They could be used for synchronization36 and encryption37 of chaotic systems.
In this paper, we propose an optimised algorithm to implement the fractional chaotic system in the numerical domain that is easy to prototype. The Jerk equation given in Refs.38,39,40,42 allow to achieve this. It will be useful to study the contribution of the fractional derivative on the dynamics of the system particularly if the implementation of the fractional system on a PIC-microcontroller allows to obtain hidden attractors contrary to the classical model considering the same parameters.
The paper is organised as follows: In the following section, we propose the description of the Jerk system followed by an analytical study. We explain the numerical methods dedicated to the computation of fractional integrals, we also present the method used for the implementation of the system on a micro-controller; “Results and discussions” is dedicated to the results and discussion; the paper ends with a conclusion and announces some perspectives for our future work.
In mechanics, a shake is a random change in the vector of acceleration without shock. In physics, the shock vector, more commonly called the jerk, is the acceleration vector’s derivative over time. Jerk systems are then the third-order differential equations of the form (frac{d^3 x}{dt^3} = F ( x)), which translates the variation of the acceleration in the system. It is the simplest of three-dimensional chaotic systems, where F(x) is the nonlinear function that describes the third-time derivative of displacement variable x. In this work, the Jerk system used will be the one presented in the Eq. (1)40.
(alpha) and (beta) could be subsequently defined as control parameters. This equation (1) can be transformed into the following system.
To solve it numerically, one use the Euler algorithm. In fact, the Euler algorithm has only one step, and it is easy to implement because it requires fewer mathematical operations41. The Euler algorithm is described for Eq. (1) by:
Physically, (x, y, z) respectively represents the position, the velocity and the acceleration (Eq. 1) could be dissipative if (beta >0). On can therefore highlight three equilibrium points, so (E_1(0,0,0)^T,; E_2(0,0,-frac{1}{sqrt{alpha }})^T), and (E_2(0,0,frac{1}{sqrt{alpha }})^T). The Jacobian matrix of (Eq. 1) is given by
We come out two characteristics equations relating to (J_0, ;J_1) and (J_2)
The literature shows that the Routh–Hurwitz criterion contains the necessary and sufficient conditions for the system stability. Thus, without even solving the characteristic equation, one finds that the system is stable for (beta >0) whatever the chosen equation.
It is shown in Refs.42,43,44,45,46,47 that the nature of the chaotic dynamic system, in addition to being sensitive to initial conditions, is closely linked to another parameter which is called the control parameter. In our case, this is the (beta) parameter. Thus, the system will behave chaotic depending on this parameter.
We report in Fig. 1a the bifurcation diagram of the system. The diagram shows a high concentration of points corresponding to the system dynamics change for (0.33<beta <0.363) and (0.371<beta <0.378). Over these intervals, the system changes periodically and it is therefore difficult to observe chaotic dynamics. In Fig. 1b, the lyapunov exponents are plotted against (beta). Indeed, Lyapunov exponents obtained using Wolf’s algorithm are another tool used to decipher the nature of a dynamical system48,49,50,51,52,53. Therefore, presence of positive Lyapunov exponents ((lambda _1)) is sufficient to establish that the considered system is able to exhibit chaotic dynamics. Moreover, the Lyapunov Exponents considered against the variation of the system’s control parameter show a large superposition with the bifurcation diagram as illustrated in the Fig. 1b. When (beta =0.35), one obtain after 1000 iteration by step of 0.01, (lambda _1= 0.089577,) (lambda _2= 0.001118 text{ and } ,lambda _3= -0.447794). The Kaplan–Yorke dimension of the system is (D_L =2.2025). Thus according to Ref.57, this Jerk system generates chaotic behaviours.
Bifurcation diagram and Lyapunov exponents for (beta in [0.328,0.43]), (alpha =frac{1}{10}), under initial conditions: ((x_0,y_0,z_0)=(0,1,0)).
There are several definitions of the fractional order derivative in the literature, but the fractional Caputo and Riemann–Liouville (R–L) operators are the most commonly used in different areas of fractional dynamical systems. The main advantage of the fractional Caputo derivative over the fractional R–L derivative is that the initial conditions of fractional differential equations with Caputo derivatives are the same as those of the integer order for differential equations53. Therefore, in this paper we will use the initial conditions ((x_0,y_0,z_0)=(0,1,0)) to solve the integer derivative and fractional derivative system. Fractional calculus is a generalization of integration and differentiation to noninteger-order fundamental operator (aD_t^q f(t))53, where a and t are the bounds of the operation. The definition of the fractional derivative under Caputo for a function f(t) of q order is defined as follows54
for (n-1< q < n) , (Gamma (.)) is Euler’s Gamma function.
Thus, system (Eq. 1) can be rewritten as (Eq. 8) to obtain the fractional Jerk system.
Hoda et al. have shown that by using Euler’s method, it is possible to find numerical solutions of linear and nonlinear systems of fractional differential equations. To prove this, they consider fractional derivatives as defined by Caputo. Furthermore, they show that Euler’s algorithm is very simple to implement and provides directly the solutions without linearization. Some examples illustrating numerical comparisons between the Euler algorithm and the classical algorithm are presented in Ref.41 to find the solution to a given dynamical system.
Let us consider the following problem: (D^qy(t)=f(t,y(t)),y(0)=y_0), (0<q<1,t>0). To solve it in the interval [0, a], it is necessary to construct a set of points ((t_j,y(t)),y(0)=y_0) which are considered as approximate values of the solution. In order to perform this approximation, the interval [0, a] is divided into n sub-intervals ([t_j,t_{j+1}]), each having an equal width. So, the general formula of the fractional Euler method is the following.
Observe that y(t) is an implicit system variable, the trapezoidal method is used to find it55,56. This method which consists in solving the system in two steps is called the prediction–correction approach that we use for and we obtain the following systems.
Applying this algorithm (Eq. 11) to the system (8), the solution is found in two steps as follows
In this section, the methodology developed in the previous section is applied on the Jerk system to observe the impact of the fractional order on the resolution result. Although the implementation is independent of the discussion, we visualise here at the same time the theoretical synchronisation analyse under Matlab and experimental under micro-controller in order to compare them.
We implemented Eqs. (12) and (13) under the Matlab software, and one observed the attractors that we present in Fig. 2. On this figure, we have several main remarks to draw. From first view, the bifurcation diagram presented in the Fig. 1 shows that the system presents a periodic dynamics for (beta =0.4) and a chaotic dynamics for (beta =0.35). Under the assumption (q_1=q_2=q_3=1), we observe for the two values of (beta) mentioned, a figure of periodic dynamics, Fig. 3a and a figure of chaotic dynamics, Fig. 2b respectively. This setting ((q_1=q_2=q_3=1)) is the equivalent of the classical resolute system. To prove the impact of the fractional order on the problem solution, let us consider the value of (beta) for which periodic dynamics are observed ((beta =0.4)) and let us varies the fractional orders. In this way, taking ((q_1,q_2,q_3)=(0.94,0.98,0.95)), Fig. 2c shows a double periodic dynamics. Still with ((beta =0.40)) and considering ((q_1,q_2,q_3)=(0.98,1,1)), one observes chaotic dynamics now (Fig. 2d), which was not observed for the classical case by considering ((q_1=q_2=q_3=1)), hence the interest of the fractional approach.
Phase portrait of the Fractional Jerk system showing route to chaos attractor when varied parameter (beta) (I): in (a) period-1 limit cycle for (beta =0.4), (b) a chaotic attractor for (beta =0.35). a and b are obtained with ((q_1,q_2,q_3)=(1,1,1)). (II): in (c) a period-2 limit cycle attractor for (beta =0.4) obtained with ((q_1,q_2,q_3)=(0.96,0.97,0.97)). (d) A chaotic attractor for (beta =0.40), obtained with ((q_1,q_2,q_3)=(0.98,1,1)) under the initial condition ((x,y,z)=(0,1,0)).
(a) Bifurcation diagram, (b) the largest Lyapunov exponents for (q=q_1=q_2=q_3), (0.3<q<1), (beta =0.40), (alpha =frac{1}{10}) under initial conditions: ((x_0 , y_0 , z_0 ) = (0, 1, 0)).
Attractor presented in Fig. 2b is a distinctive sign that the fractional system has a chaotic behaviour, but it would be interesting to carry out investigations to quantify this chaotic behaviour. To do this, we use the Lyapunov exponent applied to time series60. Indeed, the fraction system does not return the easy task to calculate the system’s Jacobian matrix. Thus, we consider a state vector of the system and we reconstruct the phase space. This method allows us to calculate the largest Lyapunov exponent, which gives us (LLE=1.43>)0, and shows that the system is also chaotic in fractional model. The influence of the fractional order on the system can thus be represented in a general way. For this purpose, we present in Fig. 3a the bifurcation diagram and in Fig. 3b the spectrum showing the evolution of the largest lyapunov exponent according to the fractional order q. The superposition observed between these two figures shows that the system dynamics is indeed influenced by the fractional order q.
The Jerk system generates real continuous values, which are not understandable by the micro-controller. In order to solve this problem, we proposed here a shift of reference frame, which allows to switch from the analogue to the digital domain. Therefore, to digitise the analogue vector x, it is necessary to know the minimum and maximum values of x, the minimum and maximum reference values of the micro-controller. This implies knowledge of the precise number of bits on which to encode the converted values. Subsequently, a linear approximation line (X(x)=ax+b) is defined, allowing to leave the interval ([x_{min},; x_{max}]) to ([X_{min},;X_{max}]). Here, (X_{min}=0), (X_{max}=(2^n-1)) and the number of bits is (n=8). So (a=frac{(2^n-1)}{x_{max}-x_{min}}) and (b=(2^n-1)x_{min}). In PICF877A micro-controller, system (3) is implemented via the Microchip XC8 compiler, the x(k) and y(k) variables was directed to the PORTB and PORTD and converted to an analogue voltage by the R-2R DAC as depicted on Fig. 4.
Schematic bloc for (x(t)-y(t)) variable.
The digital circuit hardware platform implemented for the fractional Jerk system (b) and the induced time evolution acquired by the digital oscilloscope (a).
Phase portrait of x versus y obtained experimentally of the Fractional Jerk system showing route to chaos attractor when varied parameter (beta) (A) in (a) period-1 limit cycle for (beta =0.4), (b) a chaotic attractor for (beta =0.35). a and b are obtained with ((q_1,q_2,q_3)=(1,1,1)). (B) in (c) a period-2 limit cycle attractor for (beta =0.4) obtained with ((q_1,q_2,q_3)=(0.96,0.97,0.97)). (d) a chaotic attractor for (beta =0.40), obtained with ((q_1,q_2,q_3)=(0.98,1,1)).
A Fractional Capacitor is a combination of several classical capacitors and resistors. Depending on the design method chosen, the number of resistors and capacitors used is large58. As a result, a large amount of energy is consumed. We avoid the space requirement of the device because, for an analog implementation of a third order system, three fractional capacitors need to be designed. To confirm the simplicity of a PIC implementation, observe (Fig. 5b). One notices that the experimental device contains only four main elements. There are two digital-to-analog converters (A and B), a micro-controller (C) and an oscillator of 20MHz (D). When the assembly is supplied with a 5 V voltage, oscillations are observed on the oscilloscope as shown in the Fig. 5a.
As the numerical values, we assume the same control parameter values in agreement with the bifurcation diagram. By considering the classical case ((q_1=q_2=q_3=1)) with (beta =0.35), one observes a periodic dynamics, see Fig. 6a and a chaotic dynamics, see Fig. 6b respectively which are the similar attractors to those obtained under Matlab, see Fig. 3a,b ) . To prove the impact of the fractional order on the experimental solution of the problem, we proceed as in numerical simulation, i.e. by considering the value of (beta) for which a periodic dynamics is observed ((beta =0.4)) and we vary the fractional orders. Thus, taking ((q_1,q_2,q_3)=(0.94,0.98,0.95)), the Fig. 6c shows a double periodic dynamics. Still with ((beta =0.40)) and considering ((q_1,q_2,q_3)=(0.98,1,1)), one observes a chaotic dynamics (Fig. 6d), which was not observed for the classical case considering ((q_1=q_2=q_3=1)). The same behaviour was observed in numerical simulation under Matlab (see Fig. 6c,d), which shows that the system implemented under micro-controller is successful.
Chaos synchronization consists of oscillating two chaotic systems in a synchronized manner. So, one recognize weaker forms of synchronization, when some key characteristics of the dynamical behavior are identical, such as frequencies or amplitude. Hence, in this section, we discuss the synchronization of two fractional chaotic systems. To do this, we consider here two chaotic systems called respectively master (m) and slave (s). According to the Adaptive control method59, we derive the following equations:
wherein (u_1, , u_2) and (u_3) are active non-linear controls that have been added to the chaotic chaotic system (Eq. 14) to implement the synchronisation. Considering the synchronization errors as (e_1 = x_s-x_m,~ e_2 = y_s-y_m,~ e_3 = z_s-z_m,) we derive (Eq. 16).
Equations (14) and (15) are solved using the predictor-corrector method described in “Results and discussions”. It appears that, master and slave systems trajectories converge after few milliseconds as depicted on Fig. 7 where (e_1) and (e_2) are synchronisation error, (x_m) is synchronized with (x_s). Figure 7a,b are the results obtained under Matlab, Fig. 7c,d are the synchronization results obtained experimentally.
Fractionnal synchronization result: (e_1) and (e_2) are synchronisation error (b,d), (x_m) is synchronized with (x_s) (a,c). a and b are the results obtained under Matlab, c and d are the synchronization results obtained experimentally for ((q_1 , q_2 , q_3 ) = (0.98, 1, 1).)
In this paper, we have proposed Pic micro-controller modelling a Jerk equation in integer and fractional order domains and phase portraits were investigated numerically and experimentally. Analytic studies, Lyapunov exponents and bifurcation analysis showed that the system has three determined equilibrium points and also displays complex self-excited non-linear dynamics. It appeared from simulations and experimentations that, the fractional model of the designed circuit allows to obtain masked attractors contrarily to the classical model considering the same parameters. A study case of synchronization to overcome the extreme sensitivity of the initial conditions was investigated. As a future outcome will be the exploration under a digital development board such as the FPGA.
The data that support the findings of this study are available from the corresponding author upon reasonable request.
José Francisco, G.-A., Juan, R.-G., Manuel, G.-C., José Roberto, R.-H. Fractional RC and LC electrical circuits. Ingenierí a Investigació n y Tecnologí a. 15(2), 311–319. https://doi.org/10.1016/S1405-7743(14)72219-X. (2014) (ISSN 1405-7743).
Riewe, F. Mechanics with fractional derivatives. Phys. Rev. E. 55, 3581–3592. https://doi.org/10.1103/PhysRevE.55.3581 (1997).
ADS  MathSciNet  CAS  Article  Google Scholar 
Hammad, M. M., Yaqut, A. S. H., Abdel-Khalek, M. A. & Doma, S. B. Analytical study of conformable fractional Bohr Hamiltonian with Kratzer potential. Nucl. Phys. A. 1015, 122307. https://doi.org/10.1016/j.nuclphysa.2021.122307 (2021) (ISSN 0375-9474).
CAS  Article  Google Scholar 
J. (Eds.) Kumar, S. Fractional Calculus in Medical and Health Science, 1st ed. https://doi.org/10.1201/9780429340567 (CRC Press, 2020).
Chu, Y.-M., Hadi, S.-B., Oumate, J., Abba, A., Solis-Pé rez, J. E., Bekiros, S., Gomez-Aguilar, J. F., Yousefpour, A. The effect of market confidence on a financial system from the perspective of fractional calculus: Numerical investigation and circuit realization. Chaos Solut. Fractals (2020).
Lu, H., Wu, X., Lu, J. & Wong, S. Suppression and Generation of Chaos for a Three-Dimensional Autonomous System Using Parametric Perturbations 811–819 (Cambridge University Press, 2007).
Stan, C., Cristescu, C. & Alexandroaei, D. Chaos and hyperchaos in a symmetrical discharge plasma: Experiment and modelling. UPB Sci. Bull. Ser. A Appl. Math. Phys. 70, 01 (2008).
Google Scholar 
Edward, N. L. Deterministic nonperiodic flow. J. Atmos. Sci. 20(2), 130–141, 03 (1963).
MathSciNet  Article  Google Scholar 
Vaidyanathan S. & Azar A. T. Analysis and control of a 4-D novel hyperchaotic system. In Chaos Modeling and Control Systems Design. Studies in Computational Intelligence, vol. 581 (ed. Azar A. & Vaidyanathan S.) (Springer, 2015). https://doi.org/10.1007/978-3-319-13132-0_1.
Yu, X. & Chen, G. Chaos Control: Theory and Applications (Springer, 2003).
MATH  Google Scholar 
Sandubete, J. & Escot, L. DChaos: Chaotic Time Series Analysis. R package version 0.1-5 (last version, May 2020) https://CRAN.R-project.org/package=DChaos (2019).
Fülöp, A. & Horváth, A. Application of cellular neural networks in semantic segmentation. In 2021 IEEE International Symposium on Circuits and Systems (ISCAS) 1–5. https://doi.org/10.1109/ISCAS51556.2021.9401249 (2021).
Correia, J. A. F. O. & Chang, S.-C. Controlling chaos through period-doubling bifurcations in attitude dynamics for power systems. https://doi.org/10.1155/2020/8853459 (2020).
Lynn, M. R. Edward n lorenz’s 1963 paper, “deterministic nonperiodic flow”, in journal of the atmospheric sciences, vol 20, pages 130–141: Its history and relevance to physical geography. Progr. Phys. Geogr. Earth Environ. 40(1), 175–180 (2016).
Article  Google Scholar 
Viet, P., Christos, V., Sajad, J., Xiong, W. & Tomasz, K. A simple chaotic circuit with a light-emitting diode. Communications 10, 640–646 (2016).
Google Scholar 
Hasler, M. J. Electrical circuits with chaotic behavior. Proc. IEEE 75(8), 1009–1021 (1987).
ADS  Article  Google Scholar 
Lakshmanan, M. & Murali, K. Experimental chaos from nonautonomous electronic circuits. Philos. Trans. Phys. Sci. Eng. 353(1701), 33–46 (1995).
ADS  MATH  Google Scholar 
Wang, S.-P., Lao, S.-K., Chen, H.-K., Chen, J.-H. & Chen, S.-Y. Implementation of the fractional-order Chen-Lee system by electronic circuit. Int. J. Bifurc. Chaos. 23, 1350030 (2013).
MathSciNet  Article  Google Scholar 
Chen, G. & Lü, J. A new chaotic attractor coined. Int. J. Bifurc. Chaos 12, 659–661 (2002).
MathSciNet  Article  Google Scholar 
Cafagna, D. & Grassi, G. Fractional-order Chua’s circuit: Timedomain analysis, bifurcation, chaotic behavior and test for chaos. Int. J. Bifurc. Chaos 18, 615–639 (2008).
Article  Google Scholar 
Rössler, O. E. An equation for continuous chaos. Phys. Lett. A 57, 397–398 (1976).
ADS  Article  Google Scholar 
Mekkaoui Hammouch, Z. Circuit design and simulation for the fractional-order chaotic behavior in a new dynamical system. Complex Intell. Syst. Int. J. Bifurc. Chaos. 18, 251–260 (2018).
Article  Google Scholar 
Sachin, B. Chaos in fractional order systems, 53-74. 01 (2013).
Ruan, J., Sun, K., Mou, J., He, S. & Zhang, L. Fractional-order simplest memristor-based chaotic circuit with new derivative. Eur. Phys. J. Plus 133, 01 (2018).
Article  Google Scholar 
Masoumeh, F., Bashir, N. & Yousef, E. T. The consensus of different fractional-order chaotic multiagent systems using adaptive protocols. J. Math. 2022, 5129072. https://doi.org/10.1155/2022/5129072 (2022).
MathSciNet  Article  Google Scholar 
Debnath, L. Recent applications of fractional calculus to science and engineering. Int. J. Math. Math. Sci.https://doi.org/10.1155/S0161171203301486 (2003).
MathSciNet  Article  MATH  Google Scholar 
Muira, Y., Sugi, M., Hirano, Y. & Saito, K. Simulation of fractal immittance by analog circuits: An approach to the optimized circuits. Fundam. Electron. Commun. Comput. Sci. E82–A(8), 1627–1635 (1982).
Google Scholar 
Ramesh Babu, N., Balasubramaniam, P. & Ratnavelu, K. Stability analysis of a stochastic fractional order band pass filter circuit system. AIP Conf. Proc. 2319, 060002. https://doi.org/10.1063/5.0037708 (2021).
Article  Google Scholar 
Radwan, A. G. & Salama, K. N. Fractional-order RC and RL circuits. Circuits Syst. Signal Process. 31(6), 1901–1915. https://doi.org/10.1007/s00034-012-9432-z (2012).
MathSciNet  Article  Google Scholar 
Sen, S., Tripathy, M. C. & Biswas, K. A design example of a fractional order Kerwin–Huelsman–Newcomb (KHN) biquad filter with two fractional capacitors of different order. Circuits. 32, 1523–36 (2010).
Google Scholar 
Roy, S. C. D. On the realization of a constant-argument immittance or fractional operator. IEEE Trans. Circuit Theory 14, 264–274 (1967).
Article  Google Scholar 
Carlson, G. & Halijak, C. Approximation of fractional capacitors ((1/s)^{(1/n)}) by a regular newton process. IEEE Trans. Circuit Theory 11(2), 210–213. https://doi.org/10.1109/TCT.1964.1082270 (1964).
Article  Google Scholar 
Scavo, T. R. & Thoo, J. B. On the geometry of Halley’s method. Am. Math. Monthly 102(5), 417–426 (1995).
MathSciNet  Article  Google Scholar 
Jiang, Y., Zhang, B., Shu, X. & Wei, Z. Fractional-order autonomous circuits with order larger than one. J. Adv. Res. 25, 217–225. https://doi.org/10.1016/j.jare.2020.05.005 (2020) (ISSN 2090-1232).
Article  PubMed  PubMed Central  Google Scholar 
Mouelas, A. et al. Extremely rich dynamical behaviors in a simple nonautonomous jerk system with generalized nonlinearity: Hyperchaos, intermittency, offset-boosting and multistability. Int. J. Dyn. Control. 8, 03 (2020).
Google Scholar 
Vaidyanathan, S., Takougang Kingni, S., Sambas, A., Mohamed, M. A. & Mamat, M. A new chaotic jerk system with three nonlinearities and synchronization via adaptive backstepping control. Int. J. Eng. Technol. 7, 1936–1943 (2018).
Article  Google Scholar 
Vaidyanathan, Sundarapandian, Sambas, Aceng, Mamat, Mustafa & Mada Sanjaya, W. S. Analysis, synchronisation and circuit implementation of a novel jerk chaotic system and its application for voice encryption. Int. J. Modell. Identif. Control. 28, 153 (2017).
Article  Google Scholar 
Sprott, J. C. Chaos and Time-Series Analysis. USA (2003).
Sprott, J. C. Elegant Chaos: Algebraically Simple Chaotic Flows (World Scientific, 2010).
Book  Google Scholar 
Jafari, S., Sprott, J. C. & Golpayegani, S. M. Elementary quadratic chaotic flows with no equilibria. Phys. Lett. A 377, 699–702 (2013).
ADS  MathSciNet  CAS  Article  Google Scholar 
Ahmed, H. Fractional Euler method; An effective tool for solving fractional differential equations. J. Egypt. Math. Soc. 26, 38–43. https://doi.org/10.21608/JOEMS.2018.9460 (2018).
MathSciNet  Article  MATH  Google Scholar 
Sprott, J. C. How common is chaos. Phys. Lett. A. 173(1), 21–24 (1993).
ADS  MathSciNet  Article  Google Scholar 
Sprott, J. C. Simple chaotic systems and circuits. Am. J. Phys. 68, 758. https://doi.org/10.1119/1.19538 (2000).
ADS  Article  Google Scholar 
Guckenheimer, J. H. Dynamical Systems, and Bifurcations of Vector Fields (Springer, 1983).
Book  Google Scholar 
Feckan, M. Topological Degree Approach to Bifurcation Problems (Springer, 2008).
Book  Google Scholar 
Valenza, G., Citi, L. & Barbieri, R. Estimation of instantaneous complex dynamics through Lyapunov exponents: A study on heartbeat dynamics. PLoS One 9, e105622 (2014).
ADS  Article  Google Scholar 
Wolf, A., Swift, J. B., Swinney, H. L. & Vastano, J. A. Determining Lyapunov exponents from a time series. Phys. D Nonlinear Phenom. 16(3), 285–317 (1985).
ADS  MathSciNet  Article  Google Scholar 
Sayed, W. S., Radwan, A. G., Abd-El-Hafiz, S. K. Chapter 17—Chaotic properties of various types of hidden attractors in integer and fractional order domains. In Mathematical Techniques of Fractional Order Systems, Advances in Nonlinear Dynamics and Chaos (ANDC) (eds. Azar, A. T., Radwan, A. G., & Vaidyanathan, S.) 503–528 (Elsevier, 2018).
Lyapunov, A. Problè me g é n é ral de la stabilit é du mouvement. Ann. Math. Stud. 17, 531–534 (1977) (Russian original Kharkow).
Google Scholar 
Lyapunov, A. M. Stability of Motion (Academic-Press, 1966).
Osedelec, V. I. A multiplicative ergodic theorem Lyapunov characteristic numbers for dynamical systems. Trans. Moscow Math. Soc. 19, 197–231 (1968).
Google Scholar 
Giorgilli, A., Benettin, G., Galgani, L. & Strelcyn, J.-M. Lyapunov characteristic exponents for smooth dynamical systems; a method for computing all of them part 1: Theory. Meccanica 15, 9–20 (1980).
ADS  Article  Google Scholar 
Carlson, G. E. & Halijah, C. A. Approximation of fractional capacitors (1-s)(1-n) by a regular newton process. IEEE Trans. Circuit Theory 3(7), 310–313 (1963).
Google Scholar 
Kilbas, A. A., Srivastava, H. M. & Trujillo, J. J. Theory and Applications of Fractional Differential Equations (Elsevier Science Inc., 2006).
MATH  Google Scholar 
Fuji, H. & Matsuda, K. H1 optimized wave-absorbing control: Analytical and experimental results. J. Guidance Control. Dyn. 16, 1146–1153 (1993).
ADS  Article  Google Scholar 
Ladaci, S. Contribution a la Commande Adaptative d’Ordre fractional, Thè se de Doctorat (Universit é Mentouri de Constantine, D é partement d’Electronique, 2007).
Sprott, J. C. & Ahmad, W. M. Chaos in fractional-order autonomous nonlinear systems. Chaos Solitons Fractals 16(2003), 339–351 (2002).
MATH  Google Scholar 
Jurisic, D., Emanovic, E., Lutovac, B. & Moschytz, G. Noise analysis of fractional-order two-integrator CCII low-pass filter using Pspice. https://doi.org/10.1109/MECO.2018.8406097 (2018).
Liao, T.-L. & Lin, S.-H. Adaptive control and synchronization of Lorenz systems. J. Frankl. Inst. 336(6), 925–937. https://doi.org/10.1016/S0016-0032(99)00010-1 (1999).
MathSciNet  Article  MATH  Google Scholar 
Parlitz, U. Estimating lyapunov exponents from time series. In Chaos Detection and Predictability. Lecture Notes in Physics, vol. 915 (eds. Skokos, C., Gottwald, G. & Laskar, J.) (Springer, 2016). https://doi.org/10.1007/978-3-662-48410-4_1.
Download references
This work was supported by the Alexander von Humboldt Foundation under Ref 3.4-CMR/1133622.
These authors contributed equally: Bodo Bertrand, Ayissi Eyebe Guy François Valery and J. S. A. Eyebe Fouda.
Physics, University of yaounde 1, Yaounde, 812, Cameroon
Samuel Tagne, Bertrand Bodo, Guy François V. Ayissi Eyebe & Jean Sire A. Eyebe Fouda
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
T.S.: Design, simulations, implementation, writing of the article.B.B.: Project manager, data interpretation, editing supervision.A.E.G.F.V.: Simulations model, writing of the article.J.S.A.E.F.: Project Manager, Supervision of editing.
Correspondence to Samuel Tagne.
Individual authors of this manuscript certify that they have NO affiliation or involvement with any organization or entity that has a financial interest (such as honorary or educational grants, participation in speakers’ bureaus, membership, employment, consulting, stock ownership or other participation, and expert testimony or patent licensing agreements), or a non-financial interest (such as personal or professional relationships, affiliations, knowledge, or beliefs) in the subject matter discussed in this manuscript.
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
Tagne, S., Bodo, B., Eyebe, G.F.V.A. et al. PIC micro-controller based synchronization of two fractional order jerk systems. Sci Rep 12, 14281 (2022). https://doi.org/10.1038/s41598-022-17029-x
Download citation
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41598-022-17029-x
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

Who is starring alongside BLACKPINK's Jennie in 'The Idol?' The full cast of The Weeknd's HBO show – We Got This Covered

 
 
Fans of K-pop icons BLACKPINK have been at the edge of their seats ever since the first teaser for The Idol was released a month ago. The upcoming HBO drama series created by Abel ‘The Weeknd’ Tesfaye, Reza Fahim, and Euphoria’s Sam Levinson has been gaining momentum for quite some time now. 
The premise of the drama will follow a “self-help guru and leader of a modern-day cult who enters into a complicated relationship with a rising pop idol.” While it is unclear what role Jennie will play, she will portray a character called Anys, and the teaser reveals she is a friend and colleague to Lily-Rose Depp’s character Jocelyn.
With the enormous fanbase Jennie has amassed as a member of the K-pop group, swarms of commenters have expressed their excitement at the idol’s Hollywood debut. However, she is just one out of many talented stars in The Idol’s ensemble cast. 

Here are all the stars to expect in the upcoming series that’s been tagged the “sleaziest love story in Hollywood.”
A post shared by The Weeknd (@theweeknd)
The Billboard chart-topping, Grammy Award-winning singer and songwriter is set to make his mark as the lead actor in The Idol. Going by his birth name in the credits, Tesfaye also serves as co-writer, co-creator, and executive producer of the drama. He will star as the aforementioned self-help guru and cult leader named Tedros. 
A post shared by Lily-Rose Depp (@lilyrose_depp)
With celebrity giants Johnny Depp and Vanessa Paradis as parents, Lily-Rose Depp has some serious acting genes. The model and actress will be making her television debut as Jocelyn, a pop star on the path to fame. Depp has starred in several high-profile films, including: The Dancer, Planetarium, and Netflix’s The King. 

A post shared by Debby Ryan (@debbyryan)
The former Disney Channel star became a household name in the late 2000s and early 2010s by starring in The Suite Life on Deck, 16 Wishes, Jessie, and Radio Rebel. She later earned critical praise for her lead role in Netflix’s Insatiable. Her decade is off to a great start too, with roles in Night Teeth, The Resort, and Spin Me Round. 
A post shared by suzanna son (@thestrawberrybutcher)
Model, actress, and singer Suzanna Son rose to prominence starring as Strawberry in the A24 dark comedy Red Rocket, led by Simon Rex and directed by Sean Baker. She will portray one of the cult members headed by Tesfaye’s character. 
A post shared by troye sivan (@troyesivan)
2022 is set to be the biggest acting year for South African-Australian star Troye Sivan. He took on the lead role in the Paramount Plus coming-of-age comedy-drama Three Months and earned significant praise for his performance. It’s unclear what role he will take on in the upcoming series, but from the looks of the teasers, he’s definitely getting into some trouble. Sivan is already an accomplished singer-songwriter and YouTuber, boasting millions of devoted fans. 
A post shared by • Moses Sumney (@moses)
Ghanaian-American musician Moses Sumney has been on the receiving end of critical acclaim for many years now. The multi-genre singer has released two full-length albums and five EPs, all which have been praised for their songwriting and complex music. He will make his television debut in The Idol. 
From the teasers, Jane Adams will portray a Hollywood music director in the upcoming drama. The Tony Award-winning actress has a long list of film and television credits to her name, most recently starring in the comedy Dog alongside Channing Tatum. 
A post shared by Dan Levy (@instadanjlevy)
After sweeping the Emmys for the final season of Schitt’s Creek, Dan Levy is on the fast track to taking over Hollywood. The actor has signed multi-year deals with both ABC Signature and Netflix to produce content across television. In the second teaser for The Idol, Levy was spotted alongside Jane Adams, indicating he is set to portray some form of behind-the-scenes crew member. 

A post shared by Tunde Adebimpe (@tadebimpe)
Tunde Adebimpe is a man of many talents, having starred in numerous films and shows over the years, as well as being the frontman of the band TV on the Radio. He is also a successful cartoonist, and was one of the animators on the popular MTV wrestling comedy, Celebrity Deathmatch.
A post shared by Eli Roth (@realeliroth)
One of Hollywood’s horror kings, Eli Roth is no stranger to the bloody and explicit, so The Idol should be a walk in the park for him. This will be the actor and director’s first television acting role of the decade, besides his ongoing miniseries Eli Roth’s History of Horror. 
A post shared by Rachel Anne Sennott (@treaclychild)
Comedian Rachel Sennott is famous for a number of interesting reasons. Not only did she coin the popular relationship term “soft launching,” she kept a lot of faces smiling through her videos during the heat of the 2020 pandemic. 
A post shared by @harinef
Actress and model Hari Nef is enjoying a bright and flourishing decade so far, with guest roles in the popular series Room 104, And Just Like That…, and The Marvelous Mrs. Maisel. She will also star as one of the Barbies in the highly anticipated film Barbie, directed by Greta Gerwig and starring Margot Robbie, Ryan Gosling, and a ton of other famous names.
Another fast rising star with a great start to her decade is Da’Vine Joy Randolph. With brilliant supporting and guest slots on high-profile productions, she will extend her portfolio in the upcoming HBO series. 
A post shared by MIKE DEAN (@therealmikedean)
One of hiphop’s most prolific producers and audio engineers, Mike Dean has worked with some of the biggest musical names of the last decade, including The Weeknd. 
A post shared by Hank Azaria (@hankazaria)
Charming and suave, Hank Azaria is one of the biggest names to be tapped for The Idol, most famous for his multiple voice roles including The Simpsons. The multiple Emmy Award-winning actor will bring his acting chops to the drama. 
Actors with recurring appearances include: Juliebeth Gonzalez, Steve Zissis, Elizabeth Berkley, Kyla Dyan, Liz Caribel Sierra, Maya Eshet, Melanie Liburd, Nico Haraga, and Anne Heche in her last role before her death. A24 will serve as one of the production companies for the series. The Idol has not been given an official release date. 

source

‹ Previous Next ›