Efficient stochastic parallel gradient descent training for on-chip optical processor

In recent years,space-division multiplexing(SDM)technology,which involves transmitting data information on multiple parallel channels for efficient capacity scaling,has been widely used in fiber and free-space optical communication sys-tems.To enable flexible data management and cope with the mixing between different channels,the integrated reconfig-urable optical processor is used for optical switching and mitigating the channel crosstalk.However,efficient online train-ing becomes intricate and challenging,particularly when dealing with a significant number of channels.Here we use the stochastic parallel gradient descent(SPGD)algorithm to configure the integrated optical processor,which has less com-putation than the traditional gradient descent(GD)algorithm.We design and fabricate a 6×6 on-chip optical processor on silicon platform to implement optical switching and descrambling assisted by the online training with the SPDG algorithm.Moreover,we apply the on-chip processor configured by the SPGD algorithm to optical communications for optical switching and efficiently mitigating the channel crosstalk in SDM systems.In comparison with the traditional GD al-gorithm,it is found that the SPGD algorithm features better performance especially when the scale of matrix is large,which means it has the potential to optimize large-scale optical matrix computation acceleration chips.

Designs of 3D Mesh and Torus Optical Network-on-Chips:Topology,Optical Router and Routing Module

As a nanometer-level interconnection,the Optical Network-on-Chip(ONoC)was proposed since it was typically characterized by low latency,high bandwidth and power efficiency. Compared with a 2-Dimensional(2D)design,the 3D integration has the higher packing density and the shorter wire length. Therefore,the 3D ONoC will have the great potential in the future. In this paper,we first discuss the existing ONoC researches,and then design mesh and torus ONoCs from the perspectives of topology,router,and routing module,with the help of 3D integration. A simulation platform is established by using OPNET to compare the performance of 2D and 3D ONoCs in terms of average delay and packet loss rate. The performance comparison between 3D mesh and 3D torus ONoCs is also conducted. The simulation results demonstrate that 3D integration has the advantage of reducing average delay and packet loss rate,and 3D torus ONoC has the better performance compared with 3D mesh solution. Finally,we summarize some future challenges with possible solutions,including microcosmic routing inside optical routers and highly-efficient traffic grooming.

MSONoC: a non-blocking optical interconnection network for inter cluster communication

Electric router is widely used for multi-core system to interconnect each other. However, with the increasing number of processor cores, the probability of communication conflict between processor cores increases, and the data delay increases dramatically. With the advent of optical router, the traditional electrical interconnection mode has changed to optical interconnection mode. In the packet switched optical interconnection network, the data communication mechanism consists of 3 processes: link establishment, data transmission and link termination, but the circuit-switched data transmission method greatly limits the utilization of resources. The number of micro-ring resonators in the on-chip large-scale optical interconnect network is an important parameter affecting the insertion loss. The proposed λ-route, GWOR, Crossbar structure has a large overall network insertion loss due to the use of many micro-ring resonators. How to use the least micro-ring resonator to realize non-blocking communication between multiple cores has been a research hotspot. In order to improve bandwidth and reduce access latency, an optical interconnection structure called multilevel switching optical network on chip(MSONoC) is proposed in this paper. The broadband micro-ring resonators(BMRs) are employed to reduce the number of micro-ring resonators(MRs) in the network, and the structure can provide the service of non-blocking point to point communication with the wavelength division multiplexing(WDM) technology. The results show that compared to λ-route, GWOR, Crossbar and the new topology structure, the number of micro-ring resonators of MSONoC are reduced by 95.5%, 95.5%, 87.5%, and 60% respectively. The insertion loss of the minimum link of new topology, mesh and MSONoC structure is 0.73 dB, 0.725 dB and 0.38 dB.

Pattern recognition in multi-synaptic photonic spiking neural networks based on a DFB-SA chip

Spiking neural networks(SNNs)utilize brain-like spatiotemporal spike encoding for simulating brain functions.Photonic SNN offers an ultrahigh speed and power efficiency platform for implementing high-performance neuromorphic computing.Here,we proposed a multi-synaptic photonic SNN,combining the modified remote supervised learning with delayweight co-training to achieve pattern classification.The impact of multi-synaptic connections and the robustness of the network were investigated through numerical simulations.In addition,the collaborative computing of algorithm and hardware was demonstrated based on a fabricated integrated distributed feedback laser with a saturable absorber(DFB-SA),where 10 different noisy digital patterns were successfully classified.A functional photonic SNN that far exceeds the scale limit of hardware integration was achieved based on time-division multiplexing,demonstrating the capability of hardware-algorithm co-computation.

Flexible orbital angular momentum mode switching in multimode fibre using an optical neural network chip

Mode-division multiplexing technology has been proposed as a crucial technique for enhancing communication capacity and alleviating growing communication demands.Optical switching,which is an essential component of optical communication systems,enables information exchange between channels.However,existing optical switching solutions are inadequate for addressing flexible information exchange among the mode channels.In this study,we introduced a flexible mode switching system in a multimode fibre based on an optical neural network chip.This system utilised the flexibility of on-chip optical neural networks along with an all-fibre orbital angular momentum(OAM)mode multiplexer-demultiplexer to achieve mode switching among the three OAM modes within a multimode fibre.The system adopted an improved gradient descent algorithm to achieve training for arbitrary 3×3 exchange matrices and ensured maximum crosstalk of less than-18.7 dB,thus enabling arbitrary inter-mode channel information exchange.The proposed optical-neural-network-based mode-switching system was experimentally validated by successfully transmitting different modulation formats across various modes.This innovative solution holds promise for providing effective optical switching in practical multimode communication networks.

光子神经网络研究进展

在数据海量化、信息化的时代,电子计算机处理系统所面临的算力和能耗等性能要求愈发严苛,传统冯·诺依曼架构存在“内存墙”和“功耗墙”瓶颈,加之摩尔定律放缓甚至失效,使得电子芯片在计算速度和功耗方面遇到极大挑战,利用光计算替代传统电子计算将是解决当前算力与功耗问题的极具潜力的途径之一。本文系统地梳理了片上集成和自由空间的光子神经网络架构与算法方面的研究进展,详细介绍了典型的研究工作,然后讨论并对比了这两种光子神经网络的优劣势,以及光子神经网络的训练策略等。最后探讨了光子神经网络面临的挑战,并对其未来发展进行了前瞻性的展望。

基于新奇物理现象的智能光子芯片

近年来,新材料结构、新理论观点、新器件架构等各个前沿方向的通力合作促使了智能光子学的迅猛发展。具有优异非线性特性的光学平台奠定了光子芯片丰富性能的基础,非厄米拓扑光子学等新理论的引入为高性能片上处理方案开辟了新的可能性,各类光学神经网络架构的新器件也为全光计算及全光大规模集成提供了有效的实现途径。基于上述背景,文章介绍了近年来对于智能光子学发展至关重要的非线性材料、非厄米拓扑光子学以及基于光学神经网络的新器件架构等领域的研究进展,并指出相关研究的发展对于未来更大规模、更高集成度、更复杂计算性能的智能光子芯片的实现将产生深远影响,在全光计算、全光信号处理、量子技术等领域有广泛的应用前景。

面向温度均衡的光片上网络映射

光片上网络具有带宽高、时延低和传输稳定等优点,然而光信号在光片上网络中进行数据传输时存在串扰噪声和插入损耗,影响通信性能和能耗,且在环境温度发生改变时,影响将进一步加大。光片上网络中的IP核映射将直接影响网络的流量分布,从而对网络的插入损耗和串扰噪声造成影响。文中提出的趋边快速映射算法以温度均衡为目标,通过将网络中的流量均匀分散到网络中以均衡网络温度。仿真结果证明:文中提出的映射算法在实现快速映射的同时,还能减少温度对网络传输能耗的影响。

“光芯屏端网”产业发展背景下高职院校专业设置建议

高职院校的主要任务是为我国实现“制造强国”目标培养新一代产业工人。结合湖北省经济发展现状,以电子信息类专业为例,探讨高职院校如何进行专业设置和优化,以适应地方经济发展的人才需求。提出根据不同专业特点,组建大小专业的融合,为传统专业赋能,提升学生就业率,满足地方高技术制造业的人才需要。

Low-depth optical neural networks

Optical neural network(ONNs)are emerging as attractive propos-als for machine-learning applications.However,the stability of ONNs decreases with the circuit depth,limiting the scalability of ONNs for practical uses.Here we demonstrate how to compress the circuit depth to scale only logarithmically in terms of the dimension of the data,leading to an exponential gain in terms of noise robustness.Our low-depth(LD)-ONN is based on an architecture,called Optical Com-puTing Of dot-Product UnitS(OCTOPUS),which can also be applied individually as a linear perceptron for solving classification problems.We present both numerical and theoretical evidence showing that LD-ONN can exhibit a significant improvement on robustness,compared with previous ONN proposals based on singular-value decomposition.

雷达光互连技术研究

主要探讨了雷达光互连技术的发展及其应用。介绍了光互连技术的发展历程以及在雷达系统中的演变历程,并按照雷达系统的层次,详细介绍了光互连技术在雷达各层次上的应用及目前国内外相关的光互连技术的最新成果。最后重点分析和研究了雷达系统各层次光互连技术的关键技术。

基于硅光子的片上光互连技术研究

随着芯片半导体工艺的发展,芯片集成度不断提高,单个芯片上所能容纳的计算核心数越来越多,使得核心间的数据移动效率成为制约处理器芯片整体性能的关键因素。光互连技术采用波导方式传输数据,信号传输的损耗低、速度快、延迟小,它通过采用波分复用(WDM)技术可以达到很高的带宽密度,有助于解决片上通信的瓶颈问题。面向未来片上高性能互连的需求,深入分析了电互连技术的现状与局限性,研究并分析了基于硅光子的光互连技术发展现状和趋势,对比了多种典型光互连架构的特点及优缺点,总结了未来硅光子互连技术需要解决的5个重要问题。

微系统三维集成技术的新发展

进入新世纪,微电子的发展进入纳电子/集成微系统时代,人们在继续发展摩尔定律的同时,创新了超越摩尔定律的微系统三维集成技术。介绍了在成像传感、光集成微系统、惯性传感微系统、射频微系统、生物微系统和逻辑微系统的三维集成技术的新发展,包含MEMS和IC的3D异构集成、具有Si插入器的SiP3D集成和异质3D集成等技术和各自相应的特点,以及在各应用领域所产生的革命性成果。还介绍了微系统三维集成中有关TSV的可靠性研究的最新进展。

新型高性能分层光片上网络

针对传统的片上电互连在带宽、时延等方面遇到的问题，提出一种基于光分组交换技术的分层簇的光片上网络（HCONoC）。设计了该网络的拓扑结构连接方式、布局方法、节点编址命名方法以及扩展方式等。为HCONoC的不同层次结构分别构建了新的无阻塞片上光路由器结构，并分析了损耗特性。根据网络拓扑的具体特点和编址方式，设计了有效的路由算法，可提高网络吞吐同时减少时延。最后，使用OPNET搭建仿真平台，对所提网络结构的时延和吞吐性能进行了仿真，结果表明该网络结构在本地流量模式下显示出更强的性能优势。

基于EOPCB的芯片网络

就电光印制电路板(EOPCB)上的芯间光互连提出了一种网络化的连接方案.本方案力图减少芯片间连接线数,消除芯片的控制线及时钟线,通过介质访问技术实现芯片间的互连.该方案假设每个芯片只有一个光收发器(VCSEL/PIN),芯片网络由电气层供电,并通过光互连通信.在假设条件下,芯片网络可采用总线型、环型或星型物理拓扑结构连接.给出了上述拓扑结构网络连接光路实现方法,分析了光路损耗对芯片接入数目的限制,并就适于网络化光互连的光电集成电路(OEIC)芯片结构及介质访问方式作了简单讨论.

一种用于通信C波段的光轨节点无源集成芯片

光轨是一种新型光通信网络结构,具有交换粒度小、带宽利用率高等优点。本文提出一种应用于通信C波段的新型光轨节点无源集成芯片,支撑1 545 nm、1 550 nm和1 555 nm三个C波段波长的通信。该新型光轨节点无源集成芯片是一种基于SOI纳米波导材料的片上微器件系统,核心器件由基于微环谐振器的解复用器和基于马赫-泽德尔干涉仪的环加强型热光光开关构成。通过理论计算和软件仿真,分别分析了解复用器和光开关的光学和通信性能,结果显示微环解复用器3个波长信道的串扰分别为22.5 dB、16.9 dB和16.3 dB;光开关的消光比分别为16.6 dB、19.7 dB和21.5 dB;插入损耗分别为0.86 dB、0.85 dB和0.68 dB,功耗约为51 mW。

基于多模微环谐振腔光芯片的温度测量装置设计

为提高微纳光学传感器的测温精度,基于多模微环谐振腔光芯片,搭建了高精度温度测量装置。装置通过采集微环谐振腔多模温度传感数据,利用人工神经网络算法有效融合多模传感信息,提高传感灵敏度。实验结果表明,多模传感方法在24.9~32.0℃的温度范围内,测量精度为±0.074℃,与单模微环谐振腔传感方法相比降低了温度探测极限。

Torus结构的芯片上光互连网络损耗分析和优化

基于硅基波导、十字状波导交叉和基于波导微环的光交换器件的损耗特性,对Torus结构的芯片上光互连网络建立了损耗模型,利用该模型来对芯片上光互连网络进行光器件级、光路由器级和网络级的损耗特性分析,同时建立芯片上光互连网络损耗自动分析系统.依据该系统可以得到不同网络规模下的最大损耗,并分别分析了基于Crossbar、Cygnus和Crux路由器的torus结构网络的损耗特性.传输损耗随着网络规模的扩展而增加,最小的传输损耗出现在M=N时.同时,可以得到采用Crux路由器构成的芯片上光互连网络的传输损耗最小,小于Cygnus构成的芯片上光互连网络约5 dB.

硅基光互连网络的研究

分析了硅基光互连技术的特点与其优势,归纳了几种用于硅基光互连的主要器件和相关性能参数。对硅基光互连网络的网络结构和交换机制进行研究分析后,指出了目前技术的主要问题,并提出了未来硅基光互连的研究应集中在光源集成、减小噪声串扰和优化网络结构等方面。

一种基于微环的新型4×4非阻塞光路由开关

光互连技术因诸多特性优于电互连而成为片上多核互连最具前景的解决方案。为了提高片上光互连网络架构的性能,采取光器件模块搭建的方法,提出了一种基于微环的新型4×4光路由开关,仅用7个微环构建的拓扑结构便实现了4个双向端口的非阻塞交换,降低了功耗和面积;波导交叉的数量减少到6个,优化了插入损耗。结果表明,该结构相对于经典结构光器件的功耗节省了约8%,光互连层的插入损耗降低了约7%。