Abstract:Recent advances in silicon photonic chips have made huge progress in optical computing owing to their flexibility in the reconfiguration of various tasks. Its deployment of neural networks serves as an alternative for mitigating the rapidly increased demand for computing resources in electronic platforms. However, it remains a formidable challenge to train the online programmable optical neural networks efficiently, being restricted by the difficulty in obtaining gradient information on a physical device when executing a gradient descent algorithm. Here, we experimentally demonstrate an efficient, physics-agnostic, and closed-loop protocol for training optical neural networks on chip. A gradient-free algorithm, that is, the genetic algorithm, is adopted. The protocol is on-chip implementable, physical agnostic (no need to rely on characterization and offline modeling), and gradient-free. The protocol works for various types of chip structures and is especially helpful to those that cannot be analytically decomposed and characterized. We confirm its viability using several practical tasks, including the crossbar switch and the <i>Iris</i> classification. Finally, by comparing our physics-agonistic and gradient-free method to the off-chip and gradient-based training methods, we demonstrate the robustness of our system to perturbations such as imperfect phase implementation and photodetection noise. Optical processors with gradient-free genetic algorithms have broad application potentials in pattern recognition, reinforcement learning, quantum computing, and realistic applications (such as facial recognition, natural language processing, and autonomous vehicles).The Supporting Information is available free of charge at <a class="ext-link" href="/doi/10.1021/acsphotonics.1c00035?goto=supporting-info">https://pubs.acs.org/doi/10.1021/acsphotonics.1c00035</a>.The statistical information of individuals in the final optimal generation; The results of realizing the crossbar switch on a four-mode optical processor following balanced design; The results of realizing a crossbar switch on an eight-mode optical processor following fast design; Programming a six-mode chip to realize a random T-matrix; Demonstrating the crossbar switch with multiple inputs; Blind classification with a GA-based method and numerical gradient-based method; GA-based training for handwriting digit classification; GA-based training for data set CIFAR-10 (<a class="ext-link" href="/doi/suppl/10.1021/acsphotonics.1c00035/suppl_file/ph1c00035_si_001.pdf">PDF</a>)This article has not yet been cited by other publications.

Efficient Training of Large-Scale Optical Neural Network Based on Evolutionary Algorithm

Efficient training and design of photonic neural network through neuroevolution

Efficient Self-Learning of Photonic Neural Network Through Neuroevolution

Efficient On-Chip Training of Optical Neural Networks Using Genetic Algorithm

Efficient Evolutionary Neural Architecture Search Based on Hybrid Search Space

A Gradient-Guided Evolutionary Approach to Training Deep Neural Networks

Training large-scale optoelectronic neural networks with dual-neuron optical-artificial learning

A Gradient-Guided Evolutionary Neural Architecture Search

Simple-Encoded evolving convolutional neural network and its application to skin disease image classification

Evolving Deep Convolutional Neural Networks for Image Classification

Limited Evaluation Evolutionary Optimization of Large Neural Networks

Efficient Evolutionary Neural Architecture Search by Modular Inheritable Crossover

Efficient Evolutionary Search of Attention Convolutional Networks Via Sampled Training and Node Inheritance

Multiobjective Visual Evolutionary Neural Network and Related Convolutional Neural Network Optimization

Neural Architecture Search using Covariance Matrix Adaptation Evolution Strategy

Surrogate-Assisted Evolution of Convolutional Neural Networks by Collaboratively Optimizing the Basic Blocks and Topologies

Temporal optical neurons for serial deep learning

Training Optronic Convolutional Neural Networks on an Optical System Through Backpropagation Algorithms.

A Memetic Algorithm for Evolving Deep Convolutional Neural Network in Image Classification.

Evolutionary Convolutional Neural Network Optimization with Cross-Tasks Transfer Strategy

A new genetic algorithm based evolutionary neural architecture search for image classification