Implicit neural representation(INR)networks break through the accuracy and resolution limitations of traditional discrete representations by modeling high-dimensional data as continuously differentiable implicit neura...Implicit neural representation(INR)networks break through the accuracy and resolution limitations of traditional discrete representations by modeling high-dimensional data as continuously differentiable implicit neural networks,enabling lossless compression and efficient reconstruction of details in a compact form.However,an optical-assisted INR network has yet to be demonstrated.INR networks require high nonlinearity,whereas implementing analog nonlinear activation in photonic neural networks is a challenge.Inspired by the inherent physical properties of modulators,we propose an optoelectronic nonlinear activation and implement it on the image reconstruction task.Simulations and experiments demonstrate that the proposed optoelectronic periodic neural network can represent images and perform image reconstruction with excellent results.This approach empowers complex image reconstruction with high-frequency details and reduces the amount of required hardware.Our method enables the development of compact,efficient optoelectronic neural networks,utilizing repeatable modular units for scalable and practical high-performance computing.It can enable scene generation and compression in biomedicine,autonomous driving,and augmented reality/virtual reality.展开更多
In this paper,a linear/nonlinear switching active disturbance rejection control(SADRC)based decoupling control approach is proposed to deal with some difficult control problems in a class of multi-input multi-output(M...In this paper,a linear/nonlinear switching active disturbance rejection control(SADRC)based decoupling control approach is proposed to deal with some difficult control problems in a class of multi-input multi-output(MIMO)systems such as multi-variables,disturbances,and coupling,etc.Firstly,the structure and parameter tuning method of SADRC is introduced into this paper.Followed on this,virtual control variables are adopted into the MIMO systems,making the systems decoupled.Then the SADRC controller is designed for every subsystem.After this,a stability analyzed method via the Lyapunov function is proposed for the whole system.Finally,some simulations are presented to demonstrate the anti-disturbance and robustness of SADRC,and results show SADRC has a potential applications in engineering practice.展开更多
Investigations into active noise control(ANC)technique have been conducted with the aim of effective control of the low-frequency noise.In practice,however,the performance of currently available ANC systems degrades d...Investigations into active noise control(ANC)technique have been conducted with the aim of effective control of the low-frequency noise.In practice,however,the performance of currently available ANC systems degrades due to the effects of nonlinearity in the primary and secondary paths,primary noise and louder speaker.This paper proposes a hybrid control structure of nonlinear ANC system to control the non-stationary noise produced by the rotating machinery on the nonlinear primary path.A fast version of ensemble empirical mode decomposition is used to decompose the non-stationary primary noise into intrinsic mode functions,which are expanded using the second-order Chebyshev nonlinear filter and then individually controlled.The convergence of the nonlinear ANC system is also discussed.Simulation results demonstrate that proposed method outperforms the FSLMS and VFXLMS algorithms with respect to noise reduction and convergence rate.展开更多
Parallel multi-thread processing in advanced intelligent processors is the core to realize high-speed and high-capacity signal processing systems.Optical neural network(ONN)has the native advantages of high paralleliz...Parallel multi-thread processing in advanced intelligent processors is the core to realize high-speed and high-capacity signal processing systems.Optical neural network(ONN)has the native advantages of high parallelization,large bandwidth,and low power consumption to meet the demand of big data.Here,we demonstrate the dual-layer ONN with Mach-Zehnder interferometer(MZI)network and nonlinear layer,while the nonlinear activation function is achieved by optical-electronic signal conversion.Two frequency components from the microcomb source carrying digit datasets are simultaneously imposed and intelligently recognized through the ONN.We successfully achieve the digit classification of different frequency components by demultiplexing the output signal and testing power distribution.Efficient parallelization feasibility with wavelength division multiplexing is demonstrated in our high-dimensional ONN.This work provides a high-performance architecture for future parallel high-capacity optical analog computing.展开更多
The characteristics of N-type accumulation-mode MOS (NMOS) varactors line periodically loaded with resonant tunneling diodes (RTDs) are used for soliton-like pulses generation and shaping. The problem of wide pulse br...The characteristics of N-type accumulation-mode MOS (NMOS) varactors line periodically loaded with resonant tunneling diodes (RTDs) are used for soliton-like pulses generation and shaping. The problem of wide pulse breaking up into multiple pulses rather than a single is solved. Applying perturbative analysis, we show that the dynamics of the nonlinear transmission line (NLTL) is reduced to expanded Korteweg-de Vries (KdV) equation. Moreover, numerical integration of nonlinear differential and difference equations that result from the mathematical analysis of the line is discussed. As results, NLTL can simultaneously sharpen both leading and trailing of pulse edges and one could obtain a rising and sharpening step pulse.展开更多
The rapid growth of deep learning applications has sparked a revolution in computing paradigms,with optical neural networks(ONNs)emerging as a promising platform for achieving ultra-high computing power and energy eff...The rapid growth of deep learning applications has sparked a revolution in computing paradigms,with optical neural networks(ONNs)emerging as a promising platform for achieving ultra-high computing power and energy efficiency.Despite great progress in analog optical computing,the lack of scalable optical nonlinearities and losses in photonic devices pose considerable challenges for power levels,energy efficiency,and signal latency.Here,we report an end-to-end all-optical nonlinear activator that utilizes the energy conversion of Brillouin scattering to perform efficient nonlinear processing.The activator exhibits an ultra-low activation threshold(24 nW),a wide transmission bandwidth(over 40 GHz),strong robustness,and high energy transfer efficiency.These advantages provide a feasible solution to overcome the existing bottlenecks in ONNs.As a proof-of-concept,a series of tasks is designed to validate the capability of the proposed activator as an activation unit for ONNs.Simulations show that the experiment-based nonlinear model outperforms classical activation functions in classification(97.64%accuracy for MNIST and 87.84%for Fashion-MNIST)and regression(with a symbol error rate as low as 0%)tasks.This work provides valuable insights into the innovative design of all-optical neural networks.展开更多
The classical theory of mass-spring-damper-type dynamical systems on the ordinary flat space R^3 may be generalized to higher-dimensional Riemannian manifolds by reformulating the basic underlying physical principles ...The classical theory of mass-spring-damper-type dynamical systems on the ordinary flat space R^3 may be generalized to higher-dimensional Riemannian manifolds by reformulating the basic underlying physical principles through differential geometry.Nonlinear dynamical systems have been studied in the scientific literature because they arise naturally from the modeling of complex physical structures and because such dynamical systems constitute the basis for several modern applications such as the secure transmission of information.The flows of nonlinear dynamical systems may evolve over time in complex,non-repeating(although deterministic) patterns.The focus of the present paper is on formulating the general equations that describe the dynamics of a point-wise particle sliding on a Riemannian manifold in a coordinate-free manner.The paper shows how the equations particularize in the case of some manifolds of interest in the scientific literature,such as the Stiefel manifold and the manifold of symmetric positive-definite matrices.展开更多
Optical neural networks are emerging as a competitive alternative to their electronic counterparts,offering distinct advantages in bandwidth and energy efficiency.Despite these benefits,scaling up on-chip optical neur...Optical neural networks are emerging as a competitive alternative to their electronic counterparts,offering distinct advantages in bandwidth and energy efficiency.Despite these benefits,scaling up on-chip optical neural networks for end-to-end inference is facing significant challenges.First,network depth is constrained by the weak cascadability of optical nonlinear activation functions.Second,the input size is constrained by the scale of the optical matrix.Herein,we propose a scaling up strategy called partially coherent deep optical neural networks(PDONNs).By leveraging an on-chip nonlinear activation function based on opto-electro-opto conversion,PDONN enables network depth expansion with positive net gain.Additionally,convolutional layers achieve rapid dimensionality reduction,thereby allowing for an increase in the accommodated input size.The use of a partially coherent optical source significantly reduces reliance on narrow-linewidth laser diodes and coherent detection.Owing to their broader spectral characteristics and simpler implementation,such sources are more accessible and compatible with scalable integration.Benefiting from these innovations,we designed and fabricated a monolithically integrated optical neural network with the largest input size and the deepest network depth,comprising an input layer with a size of 64,two convolutional layers,and two fully connected layers.We successfully demonstrate end-to-end two-class classification of fashion images and four-class classification of handwritten digits with accuracies of 96%and 94%,respectively,using an in-situ training method.Notably,performance is well maintained with partially coherent illumination.This proposed architecture represents a critical step toward realizing energy-efficient,scalable,and widely accessible optical computing.展开更多
Uncertain friction is a key factor that influences the accuracy of servo system in CNC machine.In this paper,based on the principle of Active Disturbance Rejection Control(ADRC),a control method is proposed,where both...Uncertain friction is a key factor that influences the accuracy of servo system in CNC machine.In this paper,based on the principle of Active Disturbance Rejection Control(ADRC),a control method is proposed,where both the extended state observer(ESO) and the reduced order extended state observer(RESO) are used to estimate and compensate for the disturbance.The authors prove that both approaches ensure high accuracy in theory,and give the criterion for parameters selection.The authors also prove that ADRC with RESO performs better than that with ESO both in disturbance estimation and tracking error.The simulation results on CNC machine show the effectiveness and feasibility of our control approaches.展开更多
基金supported by the National Natural Science Foundation of China(Grant No.62305184)the Basic and Applied Basic Research Foundation of Guangdong Province(Grant No.2023A1515012932)+1 种基金the Science,Technology and Innovation Commission of Shenzhen Municipality(Grant No.JCYJ20241202123919027)the Science,Technology and Innovation Commission of Shenzhen Municipality(Grant No.WDZC20220818100259004).
文摘Implicit neural representation(INR)networks break through the accuracy and resolution limitations of traditional discrete representations by modeling high-dimensional data as continuously differentiable implicit neural networks,enabling lossless compression and efficient reconstruction of details in a compact form.However,an optical-assisted INR network has yet to be demonstrated.INR networks require high nonlinearity,whereas implementing analog nonlinear activation in photonic neural networks is a challenge.Inspired by the inherent physical properties of modulators,we propose an optoelectronic nonlinear activation and implement it on the image reconstruction task.Simulations and experiments demonstrate that the proposed optoelectronic periodic neural network can represent images and perform image reconstruction with excellent results.This approach empowers complex image reconstruction with high-frequency details and reduces the amount of required hardware.Our method enables the development of compact,efficient optoelectronic neural networks,utilizing repeatable modular units for scalable and practical high-performance computing.It can enable scene generation and compression in biomedicine,autonomous driving,and augmented reality/virtual reality.
基金supported by the Scientific Research Innovation Development Foundation of Army Engineering University((2019)71).
文摘In this paper,a linear/nonlinear switching active disturbance rejection control(SADRC)based decoupling control approach is proposed to deal with some difficult control problems in a class of multi-input multi-output(MIMO)systems such as multi-variables,disturbances,and coupling,etc.Firstly,the structure and parameter tuning method of SADRC is introduced into this paper.Followed on this,virtual control variables are adopted into the MIMO systems,making the systems decoupled.Then the SADRC controller is designed for every subsystem.After this,a stability analyzed method via the Lyapunov function is proposed for the whole system.Finally,some simulations are presented to demonstrate the anti-disturbance and robustness of SADRC,and results show SADRC has a potential applications in engineering practice.
基金The authors greatly acknowledge the support of the National Natural Science Foundation of China under Grants 11304019 and 11774378.
文摘Investigations into active noise control(ANC)technique have been conducted with the aim of effective control of the low-frequency noise.In practice,however,the performance of currently available ANC systems degrades due to the effects of nonlinearity in the primary and secondary paths,primary noise and louder speaker.This paper proposes a hybrid control structure of nonlinear ANC system to control the non-stationary noise produced by the rotating machinery on the nonlinear primary path.A fast version of ensemble empirical mode decomposition is used to decompose the non-stationary primary noise into intrinsic mode functions,which are expanded using the second-order Chebyshev nonlinear filter and then individually controlled.The convergence of the nonlinear ANC system is also discussed.Simulation results demonstrate that proposed method outperforms the FSLMS and VFXLMS algorithms with respect to noise reduction and convergence rate.
基金Peng Xie acknowledges the support from the China Scholarship Council(Grant no.201804910829).
文摘Parallel multi-thread processing in advanced intelligent processors is the core to realize high-speed and high-capacity signal processing systems.Optical neural network(ONN)has the native advantages of high parallelization,large bandwidth,and low power consumption to meet the demand of big data.Here,we demonstrate the dual-layer ONN with Mach-Zehnder interferometer(MZI)network and nonlinear layer,while the nonlinear activation function is achieved by optical-electronic signal conversion.Two frequency components from the microcomb source carrying digit datasets are simultaneously imposed and intelligently recognized through the ONN.We successfully achieve the digit classification of different frequency components by demultiplexing the output signal and testing power distribution.Efficient parallelization feasibility with wavelength division multiplexing is demonstrated in our high-dimensional ONN.This work provides a high-performance architecture for future parallel high-capacity optical analog computing.
文摘The characteristics of N-type accumulation-mode MOS (NMOS) varactors line periodically loaded with resonant tunneling diodes (RTDs) are used for soliton-like pulses generation and shaping. The problem of wide pulse breaking up into multiple pulses rather than a single is solved. Applying perturbative analysis, we show that the dynamics of the nonlinear transmission line (NLTL) is reduced to expanded Korteweg-de Vries (KdV) equation. Moreover, numerical integration of nonlinear differential and difference equations that result from the mathematical analysis of the line is discussed. As results, NLTL can simultaneously sharpen both leading and trailing of pulse edges and one could obtain a rising and sharpening step pulse.
基金National Key Research and Development Program of China(2021YFA1401100)Innovation Group Project of Sichuan Province(20CXTD0090)。
文摘The rapid growth of deep learning applications has sparked a revolution in computing paradigms,with optical neural networks(ONNs)emerging as a promising platform for achieving ultra-high computing power and energy efficiency.Despite great progress in analog optical computing,the lack of scalable optical nonlinearities and losses in photonic devices pose considerable challenges for power levels,energy efficiency,and signal latency.Here,we report an end-to-end all-optical nonlinear activator that utilizes the energy conversion of Brillouin scattering to perform efficient nonlinear processing.The activator exhibits an ultra-low activation threshold(24 nW),a wide transmission bandwidth(over 40 GHz),strong robustness,and high energy transfer efficiency.These advantages provide a feasible solution to overcome the existing bottlenecks in ONNs.As a proof-of-concept,a series of tasks is designed to validate the capability of the proposed activator as an activation unit for ONNs.Simulations show that the experiment-based nonlinear model outperforms classical activation functions in classification(97.64%accuracy for MNIST and 87.84%for Fashion-MNIST)and regression(with a symbol error rate as low as 0%)tasks.This work provides valuable insights into the innovative design of all-optical neural networks.
基金supported by the Grant 'Ricerca Scientifica di Ateneo(RSA-B)2014'
文摘The classical theory of mass-spring-damper-type dynamical systems on the ordinary flat space R^3 may be generalized to higher-dimensional Riemannian manifolds by reformulating the basic underlying physical principles through differential geometry.Nonlinear dynamical systems have been studied in the scientific literature because they arise naturally from the modeling of complex physical structures and because such dynamical systems constitute the basis for several modern applications such as the secure transmission of information.The flows of nonlinear dynamical systems may evolve over time in complex,non-repeating(although deterministic) patterns.The focus of the present paper is on formulating the general equations that describe the dynamics of a point-wise particle sliding on a Riemannian manifold in a coordinate-free manner.The paper shows how the equations particularize in the case of some manifolds of interest in the scientific literature,such as the Stiefel manifold and the manifold of symmetric positive-definite matrices.
基金supported by the Fundamental Research Funds for the Central Universities.
文摘Optical neural networks are emerging as a competitive alternative to their electronic counterparts,offering distinct advantages in bandwidth and energy efficiency.Despite these benefits,scaling up on-chip optical neural networks for end-to-end inference is facing significant challenges.First,network depth is constrained by the weak cascadability of optical nonlinear activation functions.Second,the input size is constrained by the scale of the optical matrix.Herein,we propose a scaling up strategy called partially coherent deep optical neural networks(PDONNs).By leveraging an on-chip nonlinear activation function based on opto-electro-opto conversion,PDONN enables network depth expansion with positive net gain.Additionally,convolutional layers achieve rapid dimensionality reduction,thereby allowing for an increase in the accommodated input size.The use of a partially coherent optical source significantly reduces reliance on narrow-linewidth laser diodes and coherent detection.Owing to their broader spectral characteristics and simpler implementation,such sources are more accessible and compatible with scalable integration.Benefiting from these innovations,we designed and fabricated a monolithically integrated optical neural network with the largest input size and the deepest network depth,comprising an input layer with a size of 64,two convolutional layers,and two fully connected layers.We successfully demonstrate end-to-end two-class classification of fashion images and four-class classification of handwritten digits with accuracies of 96%and 94%,respectively,using an in-situ training method.Notably,performance is well maintained with partially coherent illumination.This proposed architecture represents a critical step toward realizing energy-efficient,scalable,and widely accessible optical computing.
基金partially supported by the National Key Basic Research Project of China under Grant No.2011CB302400the National Basic Research Program of China under Grant No.2014CB845303the National Center for Mathematics and Interdisciplinary Sciences,Chinese Academy of Sciences
文摘Uncertain friction is a key factor that influences the accuracy of servo system in CNC machine.In this paper,based on the principle of Active Disturbance Rejection Control(ADRC),a control method is proposed,where both the extended state observer(ESO) and the reduced order extended state observer(RESO) are used to estimate and compensate for the disturbance.The authors prove that both approaches ensure high accuracy in theory,and give the criterion for parameters selection.The authors also prove that ADRC with RESO performs better than that with ESO both in disturbance estimation and tracking error.The simulation results on CNC machine show the effectiveness and feasibility of our control approaches.
