Photonic neural networks(PNNs)of sufficiently large physical dimensions and high operation accuracies are envisaged as ideal candidates for breaking the major bottlenecks in the current artificial intelligence archite...Photonic neural networks(PNNs)of sufficiently large physical dimensions and high operation accuracies are envisaged as ideal candidates for breaking the major bottlenecks in the current artificial intelligence architectures in terms of latency,energy efficiency,and computational power.To achieve this vision,it is of vital importance to scale up the PNNs while simultaneously reducing the high demand on the dimensions required by them.The underlying cause of this strategy is the enormous gap between the scales of photonic and electronic integrated circuits.Here,we demonstrate monolithically integrated optical convolutional processors on thin film lithium niobate(TFLN)that harness inherent parallelism in photonics to enable large-scale programmable convolution kernels and,in turn,greatly reduce the dimensions required by subsequent fully connected layers.Experimental validation achieves high classification accuracies of 96%(86%)on the MNIST(Fashion-MNIST)dataset and 84.6%on the AG News dataset while dramatically reducing the required subsequent fully connected layer dimensions to 196×10(from 784×10)and 175×4(from 800×4),respectively.Furthermore,our devices can be driven by commercial field-programmable gate array systems;a unique advantage in addition to their scalable channel number and kernel size.Our architecture provides a solution to build practical machine learning photonic devices.展开更多
Point-of-care diagnostics and inline quantitative phase imaging(QPI)drive the demand for portable,ultra-miniaturized,and robust optical imaging and metrology systems.We propose and demonstrate a wavefront sensor integ...Point-of-care diagnostics and inline quantitative phase imaging(QPI)drive the demand for portable,ultra-miniaturized,and robust optical imaging and metrology systems.We propose and demonstrate a wavefront sensor integrated into a photonic integrated circuit,enabling single-shot optical phase retrieval.We implemented an integrated wavefront sensor array with a spatial resolution of 17μm and a numerical aperture of 0.1.Furthermore,we experimentally demonstrated the reconstruction of wavefronts defined by Zernike polynomials,specifically the first 14 terms(Z_(1)to Z_(14)),achieving an average root mean square error below 0.07.This advancement paves the way for fully integrated,portable,and robust optical imaging systems,facilitating integrated wavefront sensors in demanding applications such as point-of-care diagnostics,endoscopy,in situ QPI,and inline surface profile measurement.展开更多
Photonic neural networks have garnered significant attention in recent years due to their ultra-high computational speed,broad bandwidth,and parallel processing capabilities.However,compared to conventional electronic...Photonic neural networks have garnered significant attention in recent years due to their ultra-high computational speed,broad bandwidth,and parallel processing capabilities.However,compared to conventional electronic nonlinear activa-tion function(NAF),progress on efficient and easily implementable optical nonlinear activation function(ONAF)was barely reported.To address this issue,we proposed a programmable,low-loss ONAF device based on a silicon micro-ring resonator capped with the Antimony selenide(Sb_(2)Se_(3))thin films,and with indium tin oxide(ITO)used as the microheater.Leveraging our self-developed phase-transformation kinetic and optical models,we successfully simulated the phase-transition behavior of Sb_(2)Se_(3)and three different ONAFs—ELU,ReLU,and radial basis function(RBF)were achieved according to discernible optical responses of proposed devices under different phase-change extents.Classification results from the Fashion MNIST dataset demonstrated that these ONAFs can be considered as appropriate substitutes for traditional NAF.This indicated the bright prospect of the proposed device for nonlinear activation function in future photonic neural networks.展开更多
基金supported by the National Natural Science Foundation of China (Grant Nos.12192251,12334014,62335019,12134001,1230441812474378)+1 种基金the Quantum Science and Technology National Science and Technology Major Project(Grant No.2021ZD0301403)the Shanghai Municipal Science and Technology Major Project (Grant No.2019SHZDZX01)。
文摘Photonic neural networks(PNNs)of sufficiently large physical dimensions and high operation accuracies are envisaged as ideal candidates for breaking the major bottlenecks in the current artificial intelligence architectures in terms of latency,energy efficiency,and computational power.To achieve this vision,it is of vital importance to scale up the PNNs while simultaneously reducing the high demand on the dimensions required by them.The underlying cause of this strategy is the enormous gap between the scales of photonic and electronic integrated circuits.Here,we demonstrate monolithically integrated optical convolutional processors on thin film lithium niobate(TFLN)that harness inherent parallelism in photonics to enable large-scale programmable convolution kernels and,in turn,greatly reduce the dimensions required by subsequent fully connected layers.Experimental validation achieves high classification accuracies of 96%(86%)on the MNIST(Fashion-MNIST)dataset and 84.6%on the AG News dataset while dramatically reducing the required subsequent fully connected layer dimensions to 196×10(from 784×10)and 175×4(from 800×4),respectively.Furthermore,our devices can be driven by commercial field-programmable gate array systems;a unique advantage in addition to their scalable channel number and kernel size.Our architecture provides a solution to build practical machine learning photonic devices.
基金supported by the National Natural Science Foundation of China(Grant Nos.52175509 and 52450158)the National Key Research and Development Program of China(Grant No.2023YFF1500900)+2 种基金the Shenzhen Fundamental Research Program(Grant No.JCYJ20220818100412027)the Guangdong-Hong Kong Technology Cooperation Funding Scheme Category C Platform(Grant No.SGDX20230116093543005)the Innovation Project of Optics Valley Laboratory(Grant No.OVL2023PY003)。
文摘Point-of-care diagnostics and inline quantitative phase imaging(QPI)drive the demand for portable,ultra-miniaturized,and robust optical imaging and metrology systems.We propose and demonstrate a wavefront sensor integrated into a photonic integrated circuit,enabling single-shot optical phase retrieval.We implemented an integrated wavefront sensor array with a spatial resolution of 17μm and a numerical aperture of 0.1.Furthermore,we experimentally demonstrated the reconstruction of wavefronts defined by Zernike polynomials,specifically the first 14 terms(Z_(1)to Z_(14)),achieving an average root mean square error below 0.07.This advancement paves the way for fully integrated,portable,and robust optical imaging systems,facilitating integrated wavefront sensors in demanding applications such as point-of-care diagnostics,endoscopy,in situ QPI,and inline surface profile measurement.
基金supported by the National Natural Science Foundation of China(Grant Nos.62104114,62404111)Natural Science Foundation of Jiangsu Province(Grant Nos.BK20240635,BZ2021031)+4 种基金Opening Project of Advanced Integrated Circuit Package and Testing Research Center of Jiangsu Province(Grant No.NTIKFJJ202303)Natural Science Foundation of the Jiangsu Higher Education Institutions of China(Grant No.24KJB510025)Natural Science Research Start-up Foundation of Recruiting Talents of Nanjing University of Posts and Telecommunications(Grant Nos.NY223157,NY223156)Natural Science Foundation of Nanjing University of Posts and Telecommunications(Grant No.NY224140)Project funded by China Postdoctoral Science Foundation(Grant No.2023M732916).
文摘Photonic neural networks have garnered significant attention in recent years due to their ultra-high computational speed,broad bandwidth,and parallel processing capabilities.However,compared to conventional electronic nonlinear activa-tion function(NAF),progress on efficient and easily implementable optical nonlinear activation function(ONAF)was barely reported.To address this issue,we proposed a programmable,low-loss ONAF device based on a silicon micro-ring resonator capped with the Antimony selenide(Sb_(2)Se_(3))thin films,and with indium tin oxide(ITO)used as the microheater.Leveraging our self-developed phase-transformation kinetic and optical models,we successfully simulated the phase-transition behavior of Sb_(2)Se_(3)and three different ONAFs—ELU,ReLU,and radial basis function(RBF)were achieved according to discernible optical responses of proposed devices under different phase-change extents.Classification results from the Fashion MNIST dataset demonstrated that these ONAFs can be considered as appropriate substitutes for traditional NAF.This indicated the bright prospect of the proposed device for nonlinear activation function in future photonic neural networks.
