In a single-pixel fast imaging setup,the data collected by the single-pixel detector needs to be processed by a computer,but the speed of the latter will affect the image reconstruction time.Here we propose two kinds ...In a single-pixel fast imaging setup,the data collected by the single-pixel detector needs to be processed by a computer,but the speed of the latter will affect the image reconstruction time.Here we propose two kinds of setups which are able to transform non-visible into visible light imaging,wherein their computing process is replaced by a camera integration mode.The image captured by the camera has a low contrast,so here we present an algorithm that can realize a high quality image in near-infrared to visible cross-waveband imaging.The scheme is verified both by simulation and in actual experiments.The setups demonstrate the great potential for single-pixel imaging and high-speed cross-waveband imaging for future practical applications.展开更多
The rapid development of artificial intelligence(AI)facilitates various applications from all areas but also poses great challenges in its hardware implementation in terms of speed and energy because of the explosive ...The rapid development of artificial intelligence(AI)facilitates various applications from all areas but also poses great challenges in its hardware implementation in terms of speed and energy because of the explosive growth of data.Optical computing provides a distinctive perspective to address this bottleneck by harnessing the unique properties of photons including broad bandwidth,low latency,and high energy efficiency.In this review,we introduce the latest developments of optical computing for different AI models,including feedforward neural networks,reservoir computing,and spiking neural networks(SNNs).Recent progress in integrated photonic devices,combined with the rise of AI,provides a great opportunity for the renaissance of optical computing in practical applications.This effort requires multidisciplinary efforts from a broad community.This review provides an overview of the state-of-the-art accomplishments in recent years,discusses the availability of current technologies,and points out various remaining challenges in different aspects to push the frontier.We anticipate that the era of large-scale integrated photonics processors will soon arrive for practical AI applications in the form of hybrid optoelectronic frameworks.展开更多
Object identification and three-dimensional reconstruction techniques are always attractive research interests in machine vision,virtual reality,augmented reality,and biomedical engineering.Optical computing metasurfa...Object identification and three-dimensional reconstruction techniques are always attractive research interests in machine vision,virtual reality,augmented reality,and biomedical engineering.Optical computing metasurface,as a two-dimensional artificial design component,has displayed the supernormal character of controlling phase,amplitude,polarization,and frequency distributions of the light beam,capable of performing mathematical operations on the input light field.Here,we propose and demonstrate an all-optical object identification technique based on optical computing metasurface,and apply it to 3D reconstruction.Unlike traditional mechanisms,this scheme reduces memory consumption in the processing of the contour surface extraction.The identification and reconstruction of experimental results from high-contrast and low-contrast objects agree well with the real objects.The exploration of the all-optical object identification and 3D reconstruction techniques provides potential applications of high efficiencies,low consumption,and compact systems.展开更多
Orbital angular momentum(OAM),emerging as an inherently high-dimensional property of photons,has boosted information capacity in optical communications.However,the potential of OAM in optical computing remains almost ...Orbital angular momentum(OAM),emerging as an inherently high-dimensional property of photons,has boosted information capacity in optical communications.However,the potential of OAM in optical computing remains almost unexplored.Here,we present a highly efficient optical computing protocol for complex vector convolution with the superposition of high-dimensional OAM eigenmodes.We used two cascaded spatial light modulators to prepare suitable OAM superpositions to encode two complex vectors.Then,a deep-learning strategy is devised to decode the complex OAM spectrum,thus accomplishing the optical convolution task.In our experiment,we succeed in demonstrating 7-,9-,and 11-dimensional complex vector convolutions,in which an average proximity better than 95%and a mean relative error<6%are achieved.Our present scheme can be extended to incorporate other degrees of freedom for a more versatile optical computing in the high-dimensional Hilbert space.展开更多
The rapid rise of artificial intelligence(AI)has catalyzed advancements across various trades and professions.Developing large-scale AI models is now widely regarded as one of the most viable approaches to achieving g...The rapid rise of artificial intelligence(AI)has catalyzed advancements across various trades and professions.Developing large-scale AI models is now widely regarded as one of the most viable approaches to achieving general-purpose intelligent agents.This pressing demand has made the development of more advanced computing accelerators an enduring goal for the rapid realization of large-scale AI models.However,as transistor scaling approaches physical limits,traditional digital electronic accelerators based on the von Neumann architecture face significant bottlenecks in energy consumption and latency.Optical computing accelerators,leveraging the high bandwidth,low latency,low heat dissipation,and high parallelism of optical devices and transmission over waveguides or free space,offer promising potential to overcome these challenges.In this paper,inspired by the generic architectures of digital electronic accelerators,we conduct a bottom-up review of the principles and applications of optical computing accelerators based on the basic element of computing accelerators–the multiply-accumulate(MAC)unit.Then,we describe how to solve matrix multiplication by composing calculator arrays from different MAC units in diverse architectures,followed by a discussion on the two main applications where optical computing accelerators are reported to have advantages over electronic computing.Finally,the challenges of optical computing and our perspective on its future development are presented.Moreover,we also survey the current state of optical computing in the industry and provide insights into the future commercialization of optical computing.展开更多
Optical and hybrid convolutional neural networks(CNNs)recently have become of increasing interest to achieve low-latency,low-power image classification,and computer-vision tasks.However,implementing optical nonlineari...Optical and hybrid convolutional neural networks(CNNs)recently have become of increasing interest to achieve low-latency,low-power image classification,and computer-vision tasks.However,implementing optical nonlinearity is challenging,and omitting the nonlinear layers in a standard CNN comes with a significant reduction in accuracy.We use knowledge distillation to compress modified AlexNet to a single linear convolutional layer and an electronic backend(two fully connected layers).We obtain comparable performance with a purely electronic CNN with five convolutional layers and three fully connected layers.We implement the convolution optically via engineering the point spread function of an inverse-designed meta-optic.Using this hybrid approach,we estimate a reduction in multiply-accumulate operations from 17M in a conventional electronic modified AlexNet to only 86 K in the hybrid compressed network enabled by the optical front end.This constitutes over 2 orders of magnitude of reduction in latency and power consumption.Furthermore,we experimentally demonstrate that the classification accuracy of the system exceeds 93%on the MNIST dataset of handwritten digits.展开更多
Diffractive optical neural networks(DONNs)have exhibited the advantages of parallelization,high speed,and low consumption.However,the existing DONNs based on free-space diffractive optical elements are bulky and unste...Diffractive optical neural networks(DONNs)have exhibited the advantages of parallelization,high speed,and low consumption.However,the existing DONNs based on free-space diffractive optical elements are bulky and unsteady.In this study,we propose a planar-waveguide integrated diffractive neural network chip architecture.The three diffractive layers are engraved on the same side of a quartz wafer.The three-layer chip is designed with 32-mm3 processing space and enables a computing speed of 3.1×109 Tera operations per second.The results show that the proposed chip achieves 73.4%experimental accuracy for the Modified National Institute of Standards and Technology database while showing the system’s robustness in a cycle test.The consistency of experiments is 88.6%,and the arithmetic mean standard deviation of the results is~4.7%.The proposed chip architecture can potentially revolutionize high-resolution optical processing tasks with high robustness.展开更多
With the advancement of artificial intelligence,optic in-sensing reservoir computing based on emerging semiconductor devices is high desirable for real-time analog signal processing.Here,we disclose a flexible optomem...With the advancement of artificial intelligence,optic in-sensing reservoir computing based on emerging semiconductor devices is high desirable for real-time analog signal processing.Here,we disclose a flexible optomemristor based on C_(27)H_(30)O_(15)/FeOx heterostructure that presents a highly sensitive to the light stimuli and artificial optic synaptic features such as short-and long-term plasticity(STP and LTP),enabling the developed optomemristor to implement complex analogy signal processing through building a real-physical dynamic-based in-sensing reservoir computing algorithm and yielding an accuracy of 94.88%for speech recognition.The charge trapping and detrapping mediated by the optic active layer of C_(27)H_(30)O_(15) that is extracted from the lotus flower is response for the positive photoconductance memory in the prepared optomemristor.This work provides a feasible organic−inorganic heterostructure as well as an optic in-sensing vision computing for an advanced optic computing system in future complex signal processing.展开更多
Optical directed logic(DL)is a novel logic operation scheme that employs electrical signals as operands to control the working states of optical switches to perform the logic functions.This review first provides an ov...Optical directed logic(DL)is a novel logic operation scheme that employs electrical signals as operands to control the working states of optical switches to perform the logic functions.This review first provides an overview of the concept and working principle of DL.The developing trends of DL computing are then discussed in detail,including the fundamental optical DL gates,combinational optical DL operations,reconfigurable logic computing,low power optical logic computing,and programmable photonic network.The concluding remarks provide an outlook on the DL future development and its impacts in optical computing.展开更多
Optical computing and optical neural network have gained increasing attention in recent years because of their potential advantages of parallel processing at the speed of light and low power consumption by comparison ...Optical computing and optical neural network have gained increasing attention in recent years because of their potential advantages of parallel processing at the speed of light and low power consumption by comparison with electronic computing.The optical implementation of the fundamental building blocks of a digital computer,i.e.logic gates,has been investigated extensively in the past few decades.Optical logic gate computing is an alternative approach to various analogue optical computing architectures.In this paper,the latest development of optical logic gate computing with different kinds of implementations is reviewed.Firstly,the basic concepts of analogue and digital computing with logic gates in the electronic and optical domains are introduced.And then a comprehensive summary of various optical logic gate schemes including spatial encoding of light field,semiconductor optical amplifiers(SOA),highly nonlinear fiber(HNLF),microscale and nanoscale waveguides,and photonic crystal structures is presented.To conclude,the formidable challenges in developing practical all-optical logic gates are analyzed and the prospects of the future are discussed.展开更多
On-chip diffractive optical neural networks(DONNs)bring the advantages of parallel processing and low energy consumption.However,an accurate representation of the optical field’s evolution in the structure cannot be ...On-chip diffractive optical neural networks(DONNs)bring the advantages of parallel processing and low energy consumption.However,an accurate representation of the optical field’s evolution in the structure cannot be provided using the previous diffraction-based analysis method.Moreover,the loss caused by the open boundaries poses challenges to applications.A multimode DONN architecture based on a more precise eigenmode analysis method is proposed.We have constructed a universal library of input,output,and metaline structures utilizing this method,and realized a multimode DONN composed of the structures from the library.On the designed multimode DONNs with only one layer of the metaline,the classification task of an Iris plants dataset is verified with an accuracy of 90%on the blind test dataset,and the performance of the one-bit binary adder task is also validated.Compared to the previous architectures,the multimode DONN exhibits a more compact design and higher energy efficiency.展开更多
The explosive growth of data and information has motivated various emerging non-von Neumann computational approaches in the More-than-Moore era.Photonics neuromorphic computing has attracted lots of attention due to t...The explosive growth of data and information has motivated various emerging non-von Neumann computational approaches in the More-than-Moore era.Photonics neuromorphic computing has attracted lots of attention due to the fascinating advantages such as high speed,wide bandwidth,and massive parallelism.Here,we offer a review on the optical neural computing in our research groups at the device and system levels.The photonics neuron and photonics synapse plasticity are presented.In addition,we introduce several optical neural computing architectures and algorithms including photonic spiking neural network,photonic convolutional neural network,photonic matrix computation,photonic reservoir computing,and photonic reinforcement learning.Finally,we summarize the major challenges faced by photonic neuromorphic computing,and propose promising solutions and perspectives.展开更多
The ultimate goal of artificial intelligence(AI)is to mimic the human brain to perform decision-making and control directly from high-dimensional sensory input.Diffractive optical networks(DONs)provide a promising sol...The ultimate goal of artificial intelligence(AI)is to mimic the human brain to perform decision-making and control directly from high-dimensional sensory input.Diffractive optical networks(DONs)provide a promising solution for implementing AI with high speed and low power-consumption.Most reported DONs focus on tasks that do not involve environmental interaction,such as object recognition and image classification.By contrast,the networks capable of decision-making and control have not been developed.Here,we propose using deep reinforcement learning to implement DONs that imitate human-level decisionmaking and control capability.Such networks,which take advantage of a residual architecture,allow finding optimal control policies through interaction with the environment and can be readily implemented with existing optical devices.The superior performance is verified using three types of classic games:tic-tac-toe,Super Mario Bros.,and Car Racing.Finally,we present an experimental demonstration of playing tic-tac-toe using the network based on a spatial light modulator.Our work represents a solid step forward in advancing DONs,which promises a fundamental shift from simple recognition or classification tasks to the high-level sensory capability of AI.It may find exciting applications in autonomous driving,intelligent robots,and intelligent manufacturing.展开更多
We utilize three parallel reservoir computers using semiconductor lasers with optical feedback and light injection to model radar probe signals with delays.Three radar probe signals are generated by driving lasers con...We utilize three parallel reservoir computers using semiconductor lasers with optical feedback and light injection to model radar probe signals with delays.Three radar probe signals are generated by driving lasers constructed by a threeelement laser array with self-feedback.The response lasers are implemented also by a three-element lase array with both delay-time feedback and optical injection,which are utilized as nonlinear nodes to realize the reservoirs.We show that each delayed radar probe signal can be predicted well and to synchronize with its corresponding trained reservoir,even when parameter mismatches exist between the response laser array and the driving laser array.Based on this,the three synchronous probe signals are utilized for ranging to three targets,respectively,using Hilbert transform.It is demonstrated that the relative errors for ranging can be very small and less than 0.6%.Our findings show that optical reservoir computing provides an effective way for applications of target ranging.展开更多
Optical reservoir computing(ORC)offers advantages,such as high computational speed,low power consumption,and high training speed,so it has become a competitive candidate for time series analysis in recent years.The cu...Optical reservoir computing(ORC)offers advantages,such as high computational speed,low power consumption,and high training speed,so it has become a competitive candidate for time series analysis in recent years.The current ORC employs single-dimensional encoding for computation,which limits input resolution and introduces extraneous information due to interactions between optical dimensions during propagation,thus constraining performance.Here,we propose complex-value encoding-based optoelectronic reservoir computing(CE-ORC),in which the amplitude and phase of the input optical field are both modulated to improve the input resolution and prevent the influence of extraneous information on computation.In addition,scale factors in the amplitude encoding can fine-tune the optical reservoir dynamics for better performance.We built a CE-ORC processing unit with an iteration rate of up to∼1.2 kHz using high-speed communication interfaces and field programmable gate arrays(FPGAs)and demonstrated the excellent performance of CE-ORC in two time series prediction tasks.In comparison with the conventional ORC for the Mackey–Glass task,CE-ORC showed a decrease in normalized mean square error by∼75%.Furthermore,we applied this method in a weather time series analysis and effectively predicted the temperature and humidity within a range of 24 h.展开更多
Fourier Ptychographic Microscopy(FPM)is a high-throughput computational optical imaging technology reported in 2013.It effectively breaks through the trade-off between high-resolution imaging and wide-field imaging.In...Fourier Ptychographic Microscopy(FPM)is a high-throughput computational optical imaging technology reported in 2013.It effectively breaks through the trade-off between high-resolution imaging and wide-field imaging.In recent years,it has been found that FPM is not only a tool to break through the trade-off between field of view and spatial resolution,but also a paradigm to break through those trade-off problems,thus attracting extensive attention.Compared with previous reviews,this review does not introduce its concept,basic principles,optical system and series of applications once again,but focuses on elaborating the three major difficulties faced by FPM technology in the process from“looking good”in the laboratory to“working well”in practical applications:mismatch between numerical model and physical reality,long reconstruction time and high computing power demand,and lack of multi-modal expansion.It introduces how to achieve key technological innovations in FPM through the dual drive of Artificial Intelligence(AI)and physics,including intelligent reconstruction algorithms introducing machine learning concepts,optical-algorithm co-design,fusion of frequency domain extrapolation methods and generative adversarial networks,multi-modal imaging schemes and data fusion enhancement,etc.,gradually solving the difficulties of FPM technology.Conversely,this review deeply considers the unique value of FPM technology in potentially feeding back to the development of“AI+optics”,such as providing AI benchmark tests under physical constraints,inspirations for the balance of computing power and bandwidth in miniaturized intelligent microscopes,and photoelectric hybrid architectures.Finally,it introduces the industrialization path and frontier directions of FPM technology,pointing out that with the promotion of the dual drive of AI and physics,it will generate a large number of industrial application case,and looks forward to the possibilities of future application scenarios and expansions,for instance,body fluid biopsy and point-of-care testing at the grassroots level represent the expansion of the growth market.展开更多
Manufacturing-robust imaging systems leveraging computational optics hold immense potential for easing manufacturing constraints and enabling the development of cost-effective,high-quality imaging solutions.However,co...Manufacturing-robust imaging systems leveraging computational optics hold immense potential for easing manufacturing constraints and enabling the development of cost-effective,high-quality imaging solutions.However,conventional approaches,which typically rely on data-driven neural networks to correct optical aberrations caused by manufacturing errors,are constrained by the lack of effective tolerance analysis methods for quantitatively evaluating manufacturing error boundaries.This limitation is crucial for further relaxing manufacturing constraints and providing practical guidance for fabrication.We propose a physics-informed design paradigm for manufacturing-robust imaging systems with computational optics,integrating a physics-informed tolerance analysis methodology for evaluating manufacturing error boundaries and a physics-informed neural network for image reconstruction.With this approach,we achieve a manufacturing-robust imaging system based on an off-axis three-mirror freeform all-aluminum design,delivering a modulation transfer function exceeding 0.34 at the Nyquist frequency(72 lp/mm)in simulation.Notably,this system requires a manufacturing precision of only 0.5λin root mean square(RMS),representing a remarkable 25-fold relaxation compared with the conventional requirement of 0.02λin RMS.Experimental validation further confirmed that the manufacturing-robust imaging system maintains excellent performance in diverse indoor and outdoor environments.Our proposed method paves the way for achieving high-quality imaging without the necessity of high manufacturing precision,enabling practical solutions that are more cost-effective and time-efficient.展开更多
Artificial neural networks with internal dynamics exhibit remarkable capability in processing information.Reservoir computing(RC)is a canonical example that features rich computing expressivity and compatibility with ...Artificial neural networks with internal dynamics exhibit remarkable capability in processing information.Reservoir computing(RC)is a canonical example that features rich computing expressivity and compatibility with physical implementations for enhanced efficiency.Recently,a new RC paradigm known as next generation reservoir computing(NGRC)further improves expressivity but compromises its physical openness,posing challenges for realizations in physical systems.Here we demonstrate optical NGRC with computations performed by light scattering through disordered media.In contrast to conventional optical RC implementations,we directly and solely drive our optical reservoir with time-delayed inputs.Much like digital NGRC that relies on polynomial features of delayed inputs,our optical reservoir also implicitly generates these polynomial features for desired functionalities.By leveraging the domain knowledge of the reservoir inputs,we show that the optical NGRC not only predicts the short-term dynamics of the low-dimensional Lorenz63 and large-scale Kuramoto-Sivashinsky chaotic time series,but also replicates their long-term ergodic properties.Optical NGRC shows superiority in shorter training length and fewer hyperparameters compared to conventional optical RC based on scattering media,while achieving better forecasting performance.Our optical NGRC framework may inspire the realization of NGRC in other physical RC systems,new applications beyond time-series processing,and the development of deep and parallel architectures broadly.展开更多
The next-generation optical network is a service oriented network,which could be delivered by utilizing the generalized multiprotocol label switching(GMPLS) based control plane to realize lots of intelligent features ...The next-generation optical network is a service oriented network,which could be delivered by utilizing the generalized multiprotocol label switching(GMPLS) based control plane to realize lots of intelligent features such as rapid provisioning,automated protection and restoration(P&R),efficient resource allocation,and support for different quality of service(QoS) requirements.In this paper,we propose a novel stateful PCE-cloud(SPC)based architecture of GMPLS optical networks for cloud services.The cloud computing technologies(e.g.virtualization and parallel computing) are applied to the construction of SPC for improving the reliability and maximizing resource utilization.The functions of SPC and GMPLS based control plane are expanded according to the features of cloud services for different QoS requirements.The architecture and detailed description of the components of SPC are provided.Different potential cooperation relationships between public stateful PCE cloud(PSPC) and region stateful PCE cloud(RSPC) are investigated.Moreover,we present the policy-enabled and constraint-based routing scheme base on the cooperation of PSPC and RSPC.Simulation results for verifying the performance of routing and control plane reliability are analyzed.展开更多
文摘In a single-pixel fast imaging setup,the data collected by the single-pixel detector needs to be processed by a computer,but the speed of the latter will affect the image reconstruction time.Here we propose two kinds of setups which are able to transform non-visible into visible light imaging,wherein their computing process is replaced by a camera integration mode.The image captured by the camera has a low contrast,so here we present an algorithm that can realize a high quality image in near-infrared to visible cross-waveband imaging.The scheme is verified both by simulation and in actual experiments.The setups demonstrate the great potential for single-pixel imaging and high-speed cross-waveband imaging for future practical applications.
基金supported by the National Natural Science Foundation of China(61927802,61722209,and 61805145)the Beijing Municipal Science and Technology Commission(Z181100003118014)+3 种基金the National Key Research and Development Program of China(2020AAA0130000)the support from the National Postdoctoral Program for Innovative TalentShuimu Tsinghua Scholar Programthe support from the Hong Kong Research Grants Council(16306220)。
文摘The rapid development of artificial intelligence(AI)facilitates various applications from all areas but also poses great challenges in its hardware implementation in terms of speed and energy because of the explosive growth of data.Optical computing provides a distinctive perspective to address this bottleneck by harnessing the unique properties of photons including broad bandwidth,low latency,and high energy efficiency.In this review,we introduce the latest developments of optical computing for different AI models,including feedforward neural networks,reservoir computing,and spiking neural networks(SNNs).Recent progress in integrated photonic devices,combined with the rise of AI,provides a great opportunity for the renaissance of optical computing in practical applications.This effort requires multidisciplinary efforts from a broad community.This review provides an overview of the state-of-the-art accomplishments in recent years,discusses the availability of current technologies,and points out various remaining challenges in different aspects to push the frontier.We anticipate that the era of large-scale integrated photonics processors will soon arrive for practical AI applications in the form of hybrid optoelectronic frameworks.
基金support from the National Natural Science Foundation of China(Grant Nos.12174097 and 12304321)the Natural Science Foundation of Hunan Province(Grant Nos.2021JJ10008 and 2023JJ40202)the Research Foundation of Education Bureau of Hunan Province(Grant No.22B0871).
文摘Object identification and three-dimensional reconstruction techniques are always attractive research interests in machine vision,virtual reality,augmented reality,and biomedical engineering.Optical computing metasurface,as a two-dimensional artificial design component,has displayed the supernormal character of controlling phase,amplitude,polarization,and frequency distributions of the light beam,capable of performing mathematical operations on the input light field.Here,we propose and demonstrate an all-optical object identification technique based on optical computing metasurface,and apply it to 3D reconstruction.Unlike traditional mechanisms,this scheme reduces memory consumption in the processing of the contour surface extraction.The identification and reconstruction of experimental results from high-contrast and low-contrast objects agree well with the real objects.The exploration of the all-optical object identification and 3D reconstruction techniques provides potential applications of high efficiencies,low consumption,and compact systems.
基金supported by the National Natural Science Foundation of China(Grant Nos.12034016,61975169,and 11904303)the Youth Innovation Fund of Xiamen(Grant No.3502Z20206045)+2 种基金the Fundamental Research Funds for the Central Universities at Xiamen University(Grant Nos.20720200074 and 20720220030)the Natural Science Foundation of Fujian Province of China(Grant No.2021J02002)and for Distinguished Young Scientists(Grant No.2015J06002)the Program for New Century Excellent Talents in University of China(Grant No.NCET-13-0495).
文摘Orbital angular momentum(OAM),emerging as an inherently high-dimensional property of photons,has boosted information capacity in optical communications.However,the potential of OAM in optical computing remains almost unexplored.Here,we present a highly efficient optical computing protocol for complex vector convolution with the superposition of high-dimensional OAM eigenmodes.We used two cascaded spatial light modulators to prepare suitable OAM superpositions to encode two complex vectors.Then,a deep-learning strategy is devised to decode the complex OAM spectrum,thus accomplishing the optical convolution task.In our experiment,we succeed in demonstrating 7-,9-,and 11-dimensional complex vector convolutions,in which an average proximity better than 95%and a mean relative error<6%are achieved.Our present scheme can be extended to incorporate other degrees of freedom for a more versatile optical computing in the high-dimensional Hilbert space.
基金supported by Shanghai Municipal Science and Technology Major Project.
文摘The rapid rise of artificial intelligence(AI)has catalyzed advancements across various trades and professions.Developing large-scale AI models is now widely regarded as one of the most viable approaches to achieving general-purpose intelligent agents.This pressing demand has made the development of more advanced computing accelerators an enduring goal for the rapid realization of large-scale AI models.However,as transistor scaling approaches physical limits,traditional digital electronic accelerators based on the von Neumann architecture face significant bottlenecks in energy consumption and latency.Optical computing accelerators,leveraging the high bandwidth,low latency,low heat dissipation,and high parallelism of optical devices and transmission over waveguides or free space,offer promising potential to overcome these challenges.In this paper,inspired by the generic architectures of digital electronic accelerators,we conduct a bottom-up review of the principles and applications of optical computing accelerators based on the basic element of computing accelerators–the multiply-accumulate(MAC)unit.Then,we describe how to solve matrix multiplication by composing calculator arrays from different MAC units in diverse architectures,followed by a discussion on the two main applications where optical computing accelerators are reported to have advantages over electronic computing.Finally,the challenges of optical computing and our perspective on its future development are presented.Moreover,we also survey the current state of optical computing in the industry and provide insights into the future commercialization of optical computing.
基金supported by the National Science Foundation(Grant Nos.NSF-ECCS-2127235 and EFRI-BRAID-2223495)Part of this work was conducted at the Washington Nanofabrication Facility/Molecular Analysis Facility,a National Nanotechnology Coordinated Infrastructure(NNCI)site at the University of Washington with partial support from the National Science Foundation(Grant Nos.NNCI-1542101 and NNCI-2025489).
文摘Optical and hybrid convolutional neural networks(CNNs)recently have become of increasing interest to achieve low-latency,low-power image classification,and computer-vision tasks.However,implementing optical nonlinearity is challenging,and omitting the nonlinear layers in a standard CNN comes with a significant reduction in accuracy.We use knowledge distillation to compress modified AlexNet to a single linear convolutional layer and an electronic backend(two fully connected layers).We obtain comparable performance with a purely electronic CNN with five convolutional layers and three fully connected layers.We implement the convolution optically via engineering the point spread function of an inverse-designed meta-optic.Using this hybrid approach,we estimate a reduction in multiply-accumulate operations from 17M in a conventional electronic modified AlexNet to only 86 K in the hybrid compressed network enabled by the optical front end.This constitutes over 2 orders of magnitude of reduction in latency and power consumption.Furthermore,we experimentally demonstrate that the classification accuracy of the system exceeds 93%on the MNIST dataset of handwritten digits.
基金supported by the National Natural Science Foundation of China(Grant Nos.62175050 and U2341245)the Fundamental Research Funds for the Central Universities(Grant No.HIT.OCEF.2024054).
文摘Diffractive optical neural networks(DONNs)have exhibited the advantages of parallelization,high speed,and low consumption.However,the existing DONNs based on free-space diffractive optical elements are bulky and unsteady.In this study,we propose a planar-waveguide integrated diffractive neural network chip architecture.The three diffractive layers are engraved on the same side of a quartz wafer.The three-layer chip is designed with 32-mm3 processing space and enables a computing speed of 3.1×109 Tera operations per second.The results show that the proposed chip achieves 73.4%experimental accuracy for the Modified National Institute of Standards and Technology database while showing the system’s robustness in a cycle test.The consistency of experiments is 88.6%,and the arithmetic mean standard deviation of the results is~4.7%.The proposed chip architecture can potentially revolutionize high-resolution optical processing tasks with high robustness.
基金supported by the Key Project of Chongqing Natural Science Foundation Joint Fund[CSTB2023NSCQ-LZX0103,(G.Z.)]Chongqing Natural Science Foundation[CSTB2024NSCQ-MSX0012,(C.L.)]+1 种基金Fundamental Research Funds for the Central Universities[SWUZLPY03,(G.Z.)]Fundamental Research Funds for the Central Universities[Swu020019,(G.Z.):SWU-XDJH202319,(G.Z.)1].
文摘With the advancement of artificial intelligence,optic in-sensing reservoir computing based on emerging semiconductor devices is high desirable for real-time analog signal processing.Here,we disclose a flexible optomemristor based on C_(27)H_(30)O_(15)/FeOx heterostructure that presents a highly sensitive to the light stimuli and artificial optic synaptic features such as short-and long-term plasticity(STP and LTP),enabling the developed optomemristor to implement complex analogy signal processing through building a real-physical dynamic-based in-sensing reservoir computing algorithm and yielding an accuracy of 94.88%for speech recognition.The charge trapping and detrapping mediated by the optic active layer of C_(27)H_(30)O_(15) that is extracted from the lotus flower is response for the positive photoconductance memory in the prepared optomemristor.This work provides a feasible organic−inorganic heterostructure as well as an optic in-sensing vision computing for an advanced optic computing system in future complex signal processing.
基金This work was partially supported by the National Key R&D Program of China(No.2019YFB2205204)the Open Project Program of Wuhan National Laboratory for Optoelectronics(No.2019WNLOKF003)+1 种基金the“Shuguang Program”supported by the Shanghai Education Development Foundation and Shanghai Municipal Education Commission,the Fundamental Research Funds for the Central Universities(No.lzujbky-2021-pd11)the National Natural Science Foundation of China(Grant Nos.61875120 and 62075091).
文摘Optical directed logic(DL)is a novel logic operation scheme that employs electrical signals as operands to control the working states of optical switches to perform the logic functions.This review first provides an overview of the concept and working principle of DL.The developing trends of DL computing are then discussed in detail,including the fundamental optical DL gates,combinational optical DL operations,reconfigurable logic computing,low power optical logic computing,and programmable photonic network.The concluding remarks provide an outlook on the DL future development and its impacts in optical computing.
基金supported by the National Key Research and Development Program of China(Grants No.2021YFA1401500)the National Natural Science Foundation of China(12022416)+3 种基金the Department of Natural Resources of Guangdong Province(No.GDNRC[2022]22)Department of Science and Technology of Guangdong Province(No.2021A0505080002)Intelligent Laser Basic Research Laboratory(No.PCL2021A14-B1)the Hong Kong Research Grants Council(16306220).
文摘Optical computing and optical neural network have gained increasing attention in recent years because of their potential advantages of parallel processing at the speed of light and low power consumption by comparison with electronic computing.The optical implementation of the fundamental building blocks of a digital computer,i.e.logic gates,has been investigated extensively in the past few decades.Optical logic gate computing is an alternative approach to various analogue optical computing architectures.In this paper,the latest development of optical logic gate computing with different kinds of implementations is reviewed.Firstly,the basic concepts of analogue and digital computing with logic gates in the electronic and optical domains are introduced.And then a comprehensive summary of various optical logic gate schemes including spatial encoding of light field,semiconductor optical amplifiers(SOA),highly nonlinear fiber(HNLF),microscale and nanoscale waveguides,and photonic crystal structures is presented.To conclude,the formidable challenges in developing practical all-optical logic gates are analyzed and the prospects of the future are discussed.
基金supported by the National Natural Science Foundation of China (Grant No.62135009)the Beijing Municipal Science and Technology Commission,Administrative Commission of Zhongguancun Science Park (Grant No.Z221100005322010).
文摘On-chip diffractive optical neural networks(DONNs)bring the advantages of parallel processing and low energy consumption.However,an accurate representation of the optical field’s evolution in the structure cannot be provided using the previous diffraction-based analysis method.Moreover,the loss caused by the open boundaries poses challenges to applications.A multimode DONN architecture based on a more precise eigenmode analysis method is proposed.We have constructed a universal library of input,output,and metaline structures utilizing this method,and realized a multimode DONN composed of the structures from the library.On the designed multimode DONNs with only one layer of the metaline,the classification task of an Iris plants dataset is verified with an accuracy of 90%on the blind test dataset,and the performance of the one-bit binary adder task is also validated.Compared to the previous architectures,the multimode DONN exhibits a more compact design and higher energy efficiency.
基金This work was supported in part by the National Outstanding Youth Science Fund Project of National Natural Science Foundation of China(62022062)the National Natural Science Foundation of China(61974177,61674119)the Fundamental Research Funds for the Central Universities.
文摘The explosive growth of data and information has motivated various emerging non-von Neumann computational approaches in the More-than-Moore era.Photonics neuromorphic computing has attracted lots of attention due to the fascinating advantages such as high speed,wide bandwidth,and massive parallelism.Here,we offer a review on the optical neural computing in our research groups at the device and system levels.The photonics neuron and photonics synapse plasticity are presented.In addition,we introduce several optical neural computing architectures and algorithms including photonic spiking neural network,photonic convolutional neural network,photonic matrix computation,photonic reservoir computing,and photonic reinforcement learning.Finally,we summarize the major challenges faced by photonic neuromorphic computing,and propose promising solutions and perspectives.
基金supported by the National Natural Science Foundation of China(Grant Nos.12064025,12264028,12364045,and 12304420)the Natural Science Foundation of Jiangxi Province(Grant Nos.20212ACB202006,20232BAB201040,and 20232BAB211025)+3 种基金the Shanghai Pujiang Program(Grant No.22PJ1402900)the Australian Research Council Discovery Project(Grant No.DP200101353)the Interdisciplinary Innovation Fund of Nanchang University(Grant No.2019-9166-27060003)the China Scholarship Council(Grant No.202008420045).
文摘The ultimate goal of artificial intelligence(AI)is to mimic the human brain to perform decision-making and control directly from high-dimensional sensory input.Diffractive optical networks(DONs)provide a promising solution for implementing AI with high speed and low power-consumption.Most reported DONs focus on tasks that do not involve environmental interaction,such as object recognition and image classification.By contrast,the networks capable of decision-making and control have not been developed.Here,we propose using deep reinforcement learning to implement DONs that imitate human-level decisionmaking and control capability.Such networks,which take advantage of a residual architecture,allow finding optimal control policies through interaction with the environment and can be readily implemented with existing optical devices.The superior performance is verified using three types of classic games:tic-tac-toe,Super Mario Bros.,and Car Racing.Finally,we present an experimental demonstration of playing tic-tac-toe using the network based on a spatial light modulator.Our work represents a solid step forward in advancing DONs,which promises a fundamental shift from simple recognition or classification tasks to the high-level sensory capability of AI.It may find exciting applications in autonomous driving,intelligent robots,and intelligent manufacturing.
基金the National Natural Science Foundation of China(Grant No.62075168)Guang Dong Basic and Applied Basic Research Foundation(Grant No.2020A1515011088)Special Project in Key Fields of Guangdong Provincial Department of Education of China(Grant No.2020ZDZX3052 and 2019KZDZX1025)。
文摘We utilize three parallel reservoir computers using semiconductor lasers with optical feedback and light injection to model radar probe signals with delays.Three radar probe signals are generated by driving lasers constructed by a threeelement laser array with self-feedback.The response lasers are implemented also by a three-element lase array with both delay-time feedback and optical injection,which are utilized as nonlinear nodes to realize the reservoirs.We show that each delayed radar probe signal can be predicted well and to synchronize with its corresponding trained reservoir,even when parameter mismatches exist between the response laser array and the driving laser array.Based on this,the three synchronous probe signals are utilized for ranging to three targets,respectively,using Hilbert transform.It is demonstrated that the relative errors for ranging can be very small and less than 0.6%.Our findings show that optical reservoir computing provides an effective way for applications of target ranging.
基金the National Natural Science Foundation of China(Grant Nos.62375171,62305208,62205189,62105203,and 62405182)the Shanghai Pujiang Program(Grant No.22PJ1407500)+4 种基金the Oceanic Interdisciplinary Program of Shanghai Jiao Tong University(SJTU)(Grant No.SL2022ZD205)the National Key Research and Development Program of China(Grant No.2022YFC2806600)the Science Foundation of Donghai Laboratory(Grant Nos.DH-2022KF01001 and DH-2022KF01005)the Startup Fund for Young Faculty at SJTU(Grant No.24X010500120)the Science and Technology Commission of Shanghai Municipality(Grant No.20DZ2220400).
文摘Optical reservoir computing(ORC)offers advantages,such as high computational speed,low power consumption,and high training speed,so it has become a competitive candidate for time series analysis in recent years.The current ORC employs single-dimensional encoding for computation,which limits input resolution and introduces extraneous information due to interactions between optical dimensions during propagation,thus constraining performance.Here,we propose complex-value encoding-based optoelectronic reservoir computing(CE-ORC),in which the amplitude and phase of the input optical field are both modulated to improve the input resolution and prevent the influence of extraneous information on computation.In addition,scale factors in the amplitude encoding can fine-tune the optical reservoir dynamics for better performance.We built a CE-ORC processing unit with an iteration rate of up to∼1.2 kHz using high-speed communication interfaces and field programmable gate arrays(FPGAs)and demonstrated the excellent performance of CE-ORC in two time series prediction tasks.In comparison with the conventional ORC for the Mackey–Glass task,CE-ORC showed a decrease in normalized mean square error by∼75%.Furthermore,we applied this method in a weather time series analysis and effectively predicted the temperature and humidity within a range of 24 h.
基金National Natural Science Foundation of China(No.12574332)the Space Optoelectronic Measurement and Perception Lab.,Beijing Institute of Control Engineering(No.LabSOMP-2023-10)Major Science and Technology Innovation Program of Xianyang City(No.L2024-ZDKJ-ZDCGZH-0021)。
文摘Fourier Ptychographic Microscopy(FPM)is a high-throughput computational optical imaging technology reported in 2013.It effectively breaks through the trade-off between high-resolution imaging and wide-field imaging.In recent years,it has been found that FPM is not only a tool to break through the trade-off between field of view and spatial resolution,but also a paradigm to break through those trade-off problems,thus attracting extensive attention.Compared with previous reviews,this review does not introduce its concept,basic principles,optical system and series of applications once again,but focuses on elaborating the three major difficulties faced by FPM technology in the process from“looking good”in the laboratory to“working well”in practical applications:mismatch between numerical model and physical reality,long reconstruction time and high computing power demand,and lack of multi-modal expansion.It introduces how to achieve key technological innovations in FPM through the dual drive of Artificial Intelligence(AI)and physics,including intelligent reconstruction algorithms introducing machine learning concepts,optical-algorithm co-design,fusion of frequency domain extrapolation methods and generative adversarial networks,multi-modal imaging schemes and data fusion enhancement,etc.,gradually solving the difficulties of FPM technology.Conversely,this review deeply considers the unique value of FPM technology in potentially feeding back to the development of“AI+optics”,such as providing AI benchmark tests under physical constraints,inspirations for the balance of computing power and bandwidth in miniaturized intelligent microscopes,and photoelectric hybrid architectures.Finally,it introduces the industrialization path and frontier directions of FPM technology,pointing out that with the promotion of the dual drive of AI and physics,it will generate a large number of industrial application case,and looks forward to the possibilities of future application scenarios and expansions,for instance,body fluid biopsy and point-of-care testing at the grassroots level represent the expansion of the growth market.
基金supported by the National Natural Science Foundation of China(Grant Nos.62192774,62105243,61925504,6201101335,62020106009,62192770,62192772,62105244,62305250,and 62322217)the Science and Technology Commission of Shanghai Municipality(Grant Nos.17JC1400800,20JC1414600,and 21JC1406100)+1 种基金the Shanghai Municipal Science and Technology Major Project(Grant No.2021SHZDZX0100)the Fundamental Research Funds for the Central Universities.
文摘Manufacturing-robust imaging systems leveraging computational optics hold immense potential for easing manufacturing constraints and enabling the development of cost-effective,high-quality imaging solutions.However,conventional approaches,which typically rely on data-driven neural networks to correct optical aberrations caused by manufacturing errors,are constrained by the lack of effective tolerance analysis methods for quantitatively evaluating manufacturing error boundaries.This limitation is crucial for further relaxing manufacturing constraints and providing practical guidance for fabrication.We propose a physics-informed design paradigm for manufacturing-robust imaging systems with computational optics,integrating a physics-informed tolerance analysis methodology for evaluating manufacturing error boundaries and a physics-informed neural network for image reconstruction.With this approach,we achieve a manufacturing-robust imaging system based on an off-axis three-mirror freeform all-aluminum design,delivering a modulation transfer function exceeding 0.34 at the Nyquist frequency(72 lp/mm)in simulation.Notably,this system requires a manufacturing precision of only 0.5λin root mean square(RMS),representing a remarkable 25-fold relaxation compared with the conventional requirement of 0.02λin RMS.Experimental validation further confirmed that the manufacturing-robust imaging system maintains excellent performance in diverse indoor and outdoor environments.Our proposed method paves the way for achieving high-quality imaging without the necessity of high manufacturing precision,enabling practical solutions that are more cost-effective and time-efficient.
基金supported by Swiss National Science Foundation(SNF)projects LION,ERC SMARTIES and Institut Universitaire de France.H.W.acknowledges China Scholarship Council and National Natural Science Foundation of China(623B2064 and 62275137)J.H.acknowledges SNF fellowship(P2ELP2_199825)+3 种基金Y.B.acknowledges the support from Basic Science Research Program through the National Research Foundation of Korea(NRF)funded by the Ministry of Education(2022R1A6A3A03072108)European Union’s Horizon Europe research and innovation program(N.101105899)Q.L.acknowledges National Natural Science Foundation of China(62275137)the Tsinghua University(Department of Precision Instrument)-North Laser Research Institute Co.,Ltd Joint Research Center for Advanced Laser Technology(20244910194).
文摘Artificial neural networks with internal dynamics exhibit remarkable capability in processing information.Reservoir computing(RC)is a canonical example that features rich computing expressivity and compatibility with physical implementations for enhanced efficiency.Recently,a new RC paradigm known as next generation reservoir computing(NGRC)further improves expressivity but compromises its physical openness,posing challenges for realizations in physical systems.Here we demonstrate optical NGRC with computations performed by light scattering through disordered media.In contrast to conventional optical RC implementations,we directly and solely drive our optical reservoir with time-delayed inputs.Much like digital NGRC that relies on polynomial features of delayed inputs,our optical reservoir also implicitly generates these polynomial features for desired functionalities.By leveraging the domain knowledge of the reservoir inputs,we show that the optical NGRC not only predicts the short-term dynamics of the low-dimensional Lorenz63 and large-scale Kuramoto-Sivashinsky chaotic time series,but also replicates their long-term ergodic properties.Optical NGRC shows superiority in shorter training length and fewer hyperparameters compared to conventional optical RC based on scattering media,while achieving better forecasting performance.Our optical NGRC framework may inspire the realization of NGRC in other physical RC systems,new applications beyond time-series processing,and the development of deep and parallel architectures broadly.
基金supported by National Natural Science Foundation of China(No.61571061)Innovative Research Fund of Beijing University of Posts and Telecommunications (2015RC16)
文摘The next-generation optical network is a service oriented network,which could be delivered by utilizing the generalized multiprotocol label switching(GMPLS) based control plane to realize lots of intelligent features such as rapid provisioning,automated protection and restoration(P&R),efficient resource allocation,and support for different quality of service(QoS) requirements.In this paper,we propose a novel stateful PCE-cloud(SPC)based architecture of GMPLS optical networks for cloud services.The cloud computing technologies(e.g.virtualization and parallel computing) are applied to the construction of SPC for improving the reliability and maximizing resource utilization.The functions of SPC and GMPLS based control plane are expanded according to the features of cloud services for different QoS requirements.The architecture and detailed description of the components of SPC are provided.Different potential cooperation relationships between public stateful PCE cloud(PSPC) and region stateful PCE cloud(RSPC) are investigated.Moreover,we present the policy-enabled and constraint-based routing scheme base on the cooperation of PSPC and RSPC.Simulation results for verifying the performance of routing and control plane reliability are analyzed.
