Deep learning with coherent nanophotonic circuits

2016/10/07 by Yichen Shen, Nicholas C. Harris, Scott A. Skirlo +11 · 2 voices · 89 citations
Computer Science · Engineering · #Neural Networks and Reservoir Computing #Photonic and Optical Devices #Optical Network Technologies

paper · pdf · doi:10.1038/nphoton.2017.93

Abstract

Artificial Neural Networks are computational network models inspired by signal processing in the brain. These models have dramatically improved the performance of many learning tasks, including speech and object recognition. However, today's computing hardware is inefficient at implementing neural networks, in large part because much of it was designed for von Neumann computing schemes. Significant effort has been made to develop electronic architectures tuned to implement artificial neural networks that improve upon both computational speed and energy efficiency. Here, we propose a new architecture for a fully-optical neural network that, using unique advantages of optics, promises a computational speed enhancement of at least two orders of magnitude over the state-of-the-art and three orders of magnitude in power efficiency for conventional learning tasks. We experimentally demonstrate essential parts of our architecture using a programmable nanophotonic processor.

Citations

• ImageNet classification with deep convolutional neural networks

Cited by

• Applications and Techniques for Fast Machine Learning in Science
• Low-Loss and Broadband Nonvolatile Phase-Change Directional Coupler Switches
• Photonic convolutional neural network with pre-trained in situ training
• HAMON: Passive Optical Sequence Mixing for Long-Horizon Forecasting
• All-optical nonlinear phase modulation in open semiconductor microcavities
• Programmable photonics enabled by ferroionic two-dimensional materials
• Backpropagation-Free Trunk Training via the Split Forward Gradients
• Local Variance-Based Calibration of Programmable Photonic Interferometer Meshes
• Continuous-variable photonic quantum extreme learning machines for fast collider-data selection
• Free‐space Optical Computing Systems
• Integrated photonic synapses, neurons, memristors, and neural networks for photonic neuromorphic computing
• Review for wireless communication technology based on digital encoding metasurfaces
• Approximate reservoir computing with a semiconductor laser for reducing energy consumption
• Reconfigurable Nonlinear Photonic Networks for In-Situ Learning and Memory Formation via Driven-Dissipative Dynamics
• End-to-end Optimization of Single-Shot Quantum Machine Learning for Bayesian Inference
• Nanophotonic Inverse Design with SPINS: Software Architecture and Practical Considerations
• Multilevel Photonic Switching in GST-467 for Deep Neural Network Inference
• A Segmented Heater-Driven, Low-Loss, Reconfigurable Photonic Phase-Change Material-Based Phase Shifter
• Reconfigurable Silicon Photonics Extreme Learning Machine with Random Non-linearities as Neural Processor and Physical Unclonable Function
• Phase-multiplexed optical computing: Reconfiguring a multi-task diffractive optical processor using illumination phase diversity
• Metasurface-based all-optical diffractive convolutional neural networks
• Near-Zero Crosstalk and Ultra-Low Loss Waveguide Crossings Enabled by three-dimensional Ta2O5-on-LNOI Integrated Photonic Platform
• An all-optical convolutional neural network for image identification
• All-optical spiking neurosynaptic networks with self-learning capabilities
• Hardware-aware Lightweight Photonic Spiking Neural Network for Pattern Classification
• A scalable advantage in multi-photon quantum machine learning
• Optical Proof of Work
• The Immutable Tensor Architecture: A Pure Dataflow Approach for Secure, Energy-Efficient AI Inference
• Dispersion-Aware Modeling Framework for Parallel Optical Computing
• Rethinking Vision Transformer Depth via Structural Reparameterization
• Multi-port programmable silicon photonics using low-loss phase change material Sb₂Se₃
• Machine learning and the physical sciences
• Photonic neural field on a silicon chip: large-scale, high-speed neuro-inspired computing and sensing
• Quantum optical neural networks using atom-cavity interactions to provide all-optical nonlinearity
• Ultrafast Reconfigurable Topological Photonic Processing Accelerator
• A Case Study on the Performance Metrics of Integrated Photonic Computing
• LightPro: A Linear Photonic Processor with Full Programmability
• Artificial Intelligence Accelerators based on Graphene Optoelectronic Devices
• Roadmap on embodying mechano-intelligence and computing in functional materials and structures
• Artificial Intelligence in Nanophotonics: From Design to Optical Computing
• Conflict-free collective stochastic decision making by orbital angular momentum of photons through quantum interference
• Architecture, Simulation and Software Stack to Support Post-CMOS Accelerators: The ARCHYTAS Project
• Scalar computational primitives with perturbative phase interferometry
• Automated control strategy for setting and stabilization of photonic circuits
• Resonant tunnelling diode nano-optoelectronic spiking nodes for\n neuromorphic information processing
• Shifted rectangular mesh architecture for programmable photonics
• Approximate Proximal Operators for Analog Compressed Sensing Using PN-junction Diode
• Nonlinear optical feature generator for machine learning
• ENLighten: Lighten the Transformer, Enable Efficient Optical Acceleration
• Electromagnetic Eigenvalue Problems and Nonhermitian Effects in Linear and Saturable Scattering
• Probing a theoretical framework for a Photonic Extreme Learning Machine
• Photonics-Aware Planning-Guided Automated Electrical Routing for Large-Scale Active Photonic Integrated Circuits
• Spatially Parallel All-optical Neural Networks
• Blending Optimal Control and Biologically Plausible Learning for Noise-Robust Physical Neural Networks
• Survey of Machine Learning Accelerators
• Design, fabrication, and metrology of 10 × 100 multi-planar integrated photonic routing manifolds for neural networks
• On the basis of brain: neural-network-inspired changes in general-purpose chips
• High Clockrate Free-space Optical In-Memory Computing
• Can AI Follow In Einstein's Footsteps?
• Energy-efficient photonic neural networks for high-speed AI computation
• Homodyne Photonic Tensor Processor exceeds 1,000-TOPS
• Fiber-based diffractive deep neural network
• Manufacturing of graphene based synaptic devices for optoelectronic applications
• Training deep physical neural networks with local physical information bottleneck
• All-optical machine learning using diffractive deep neural networks
• Heuristic recurrent algorithms for photonic Ising machines
• Wave physics as an analog recurrent neural network
• PCNNA: A Photonic Convolutional Neural Network Accelerator
• Reprogrammable Electro-Optic Nonlinear Activation Functions for Optical Neural Networks
• Topological Photon Transport in Programmable Photonic Processors via Discretized Evolution of Synthetic Magnetic Fields
• Photonic decision making using optical frequency difference detection in mutually-coupled semiconductor lasers
• Femtojoule-per-operation photonic computer for the subset sum problem
• Single photonic perceptron based on a soliton crystal Kerr microcomb for high-speed, scalable, optical neural networks
• Hybrid integration of two-dimensional dichalcogenides for low power saturable absorption in photonic integrated circuits
• Toward Lifelong-Sustainable Electronic-Photonic AI Systems via Extreme Efficiency, Reconfigurability, and Robustness
• Universality of physical neural networks with multivariate nonlinearity
• Performance evaluation of an integrated photonic convolutional neural network based on delay buffering and wavelength division multiplexing
• Integrated photonic neuromorphic computing: device, architecture, chip, algorithm
• The physics of optical computing
• Near-optimal decomposition of unitary matrices using phase masks and the discrete Fourier transform
• On-chip wave chaos for photonic extreme learning
• A fully-programmable integrated photonic processor for both domain-specific and general-purpose computing
• Programmable photonic circuits
• Incoherent Light-Driven Nonlinear Optical Extreme Learner via Data Reverberation
• Nonlinear Photonic Neuromorphic Chips for Spiking Reinforcement Learning
• Polymer fiber mats provide white and full-color tunable lasing
• Mixed Precision Photonic Computing with 3D Electronic-Photonic Integrated Circuits
• Adiabatic evolution on a spatial-photonic Ising machine
• Dirk Englund [wikipedia]

Discussions

• Deep learning with coherent nanophotonic circuits [pdf] [hn, 82 points, 9 comments]
• Deep Learning with Coherent Nanophotonic Circuits [hn, 2 points, 0 comments]

Related

• Spatially Varying Nanophotonic Neural Networks
• Optical Computing for Deep Neural Network Acceleration: Foundations, Recent Developments, and Emerging Directions
• A compact butterfly-style silicon photonic-electronic neural chip for hardware-efficient deep learning
• Photonic Neural Networks and Optics-informed Deep Learning Fundamentals
• On-Chip Optical Convolutional Neural Networks