Wave Physics as an Analog Recurrent Neural Network

Abstract

Analog machine learning hardware platforms promise to be faster and more energy-efficient than their digital counterparts. Wave physics, as found in acoustics and optics, is a natural candidate for building analog processors for time-varying signals. Here we identify a mapping between the dynamics of wave physics, and the computation in recurrent neural networks. This mapping indicates that physical wave systems can be trained to learn complex features in temporal data, using standard training techniques for neural networks. As a demonstration, we show that an inverse-designed inhomogeneous medium can perform vowel classification on raw audio signals as their waveforms scatter and propagate through it, achieving performance comparable to a standard digital implementation of a recurrent neural network. These findings pave the way for a new class of analog machine learning platforms, capable of fast and efficient processing of information in its native domain.

arXiv comment13 pages, 6 figures

References in corpus (15)

• Adam: A Method for Stochastic Optimization
• Sequence to Sequence Learning with Neural Networks
• Empirical Evaluation of Gated Recurrent Neural Networks on Sequence Modeling
• Quantum Machine Learning
• Deep Learning with Coherent Nanophotonic Circuits
• All-Optical Machine Learning Using Diffractive Deep Neural Networks
• Outlook for inverse design in nanophotonics
• Neuromorphic Silicon Photonic Networks
• Vowel recognition with four coupled spin-torque nano-oscillators
• Training of photonic neural networks through in situ backpropagation
• A Photonic Crystal Slab Laplace Differentiator
• Superconducting optoelectronic circuits for neuromorphic computing
• Tunable Efficient Unitary Neural Networks (EUNN) and their application to RNNs
• Trainable and Dynamic Computing: Error Backpropagation through Physical Media
• Nanophotonic Media for Artificial Neural Inference

Cited by in corpus (81)

• Deep physical neural networks enabled by a backpropagation algorithm for arbitrary physical systems
• Large-scale neuromorphic optoelectronic computing with a reconfigurable diffractive processing unit
• Deep neural networks for the evaluation and design of photonic devices
• Deep learning in nano-photonics: inverse design and beyond
• Inverse design of photonic crystals through automatic differentiation
• Spectrally-Encoded Single-Pixel Machine Vision Using Diffractive Networks
• Nanoscale neural network using non-linear spin-wave interference
• Neuromorphic computing using wavelength-division multiplexing
• Tailoring Structure-borne Sound Through Bandgap Engineering in Phononic Crystals and Metamaterials: A Comprehensive Review
• Large-Scale Optical Reservoir Computing for Spatiotemporal Chaotic Systems Prediction
• Theory of neuromorphic computing by waves: machine learning by rogue waves, dispersive shocks, and solitons
• Photonic tensor cores for machine learning
• Prospects and applications of photonic neural networks
• Terahertz Pulse Shaping Using Diffractive Surfaces
• Class-specific Differential Detection in Diffractive Optical Neural Networks Improves Inference Accuracy
• Deep Learning in Deterministic Computational Mechanics
• All-Optical Information Processing Capacity of Diffractive Surfaces
• Misalignment Resilient Diffractive Optical Networks
• Ensemble learning of diffractive optical networks
• Artificial Neural Network with Physical Dynamic Metasurface Layer for Optimal Sensing
• Solving integral equations in free-space with inverse-designed ultrathin optical metagratings
• Polarization Multiplexed Diffractive Computing: All-Optical Implementation of a Group of Linear Transformations Through a Polarization-Encoded Diffractive Network
• A General Approach to Seismic Inversion with Automatic Differentiation
• Photonic extreme learning machine by free-space optical propagation
• Intelligent Meta-Imagers: From Compressed to Learned Sensing
• Massively Parallel Universal Linear Transformations using a Wavelength-Multiplexed Diffractive Optical Network
• Forward-Mode Differentiation of Maxwell's Equations
• Scale-, shift- and rotation-invariant diffractive optical networks
• Roadmap for Unconventional Computing with Nanotechnology
• Artificial Neural Networks for Photonic Applications: From Algorithms to Implementation
• Arbitrary linear transformations for photons in the frequency synthetic dimension
• Deep learning enabled design of complex transmission matrices for universal optical components
• Deterministic and probabilistic deep learning models for inverse design of broadband acoustic cloak
• A single inverse-designed photonic structure that performs parallel computing
• MetaNet: A new paradigm for data sharing in photonics research
• Symplectic Neural Networks in Taylor Series Form for Hamiltonian Systems
• Inverse design of plasma metamaterial devices for optical computing
• Reservoir Computing and Photoelectrochemical Sensors: A Marriage of Convenience
• Far-field subwavelength acoustic imaging by deep learning
• Inverse Design of Grating Couplers Using the Policy Gradient Method from Reinforcement Learning
• Neural Schrödinger Equation:Physical Law as Neural Network
• Seismic Inversion by Hybrid Machine Learning
• Learning Stochastic Dynamics with Statistics-Informed Neural Network
• Energy-efficient neural network inference with microcavity exciton-polaritons
• Wave-based extreme deep learning based on non-linear time-Floquet entanglement
• Self-learning Machines based on Hamiltonian Echo Backpropagation
• Programmable wave-based analog computing machine: a metastructure that designs metastructures
• Stimulated amplification of propagating spin waves
• Quantum-limited stochastic optical neural networks operating at a few quanta per activation
• Ultrafast single-channel machine vision based on neuro-inspired photonic computing
• Mesoscopic ultrafast nonlinear optics -- The emergence of multimode quantum non-Gaussian physics
• An optoacoustic field-programmable perceptron for recurrent neural networks
• Optical Neural Network Based on Synthetic Nonlinear Photonic Lattices
• Kernel Phase and Coronagraphy with Automatic Differentiation
• Nonlinear erasing of propagating spin-wave pulses in thin-film Ga:YIG
• Complete identification of spin-wave eigenmodes excited by parametric pumping in YIG microdisks
• Artificial Nanophotonic Neuron with Internal Memory for Biologically Inspired and Reservoir Network Computing
• Optical neuromorphic computing via temporal up-sampling and trainable encoding on a telecom device platform
• Optical neural network architecture for deep learning with the temporal synthetic dimension
• Arbitrary control over multimode wave propagation for machine learning
• Physical Computing for Materials Acceleration Platforms
• Model-free Optical Processors using In Situ Reinforcement Learning with Proximal Policy Optimization
• A physics-defined recurrent neural network to compute coherent light wave scattering on the millimetre scale
• Neural Vortex Method: from Finite Lagrangian Particles to Infinite Dimensional Eulerian Dynamics
• Blending Optimal Control and Biologically Plausible Learning for Noise-Robust Physical Neural Networks
• Mechanical in-sensor computing: a programmable meta-sensor for structural damage classification without external electronic power
• A machine learning model to classify dynamic processes in liquid water
• Neural network is heterogeneous: Phase matters more
• Numerical analysis of a deep learning formulation of elastic full waveform inversion with high order total variation regularization in different parameterization
• Gaussian boson sampling and multi-particle event optimization by machine learning in the quantum phase space
• Partial Differential Equations is All You Need for Generating Neural Architectures -- A Theory for Physical Artificial Intelligence Systems
• Time-Continuous Energy-Conservation Neural Network for Structural Dynamics Analysis
• A single metagrating metastructure for wave-based parallel analog computing
• Induced Homomorphism Kirchhoffs Law in Photonics
• Tensor Network for Supervised Learning at Finite Temperature
• Wavefield solutions from machine learned functions
• Phase-multiplexed optical computing: Reconfiguring a multi-task diffractive optical processor using illumination phase diversity
• Oscillator-Based Processing Unit for Formant Recognition
• Compatible Learning for Deep Photonic Neural Network
• Nonseparable Symplectic Neural Networks
• $\mathbb{T}$-Operator Limits on Optical Communication: Metaoptics, Computation, and Input-Output Transformations