Abstract:Optical computing promises to improve the speed and energy efficiency of machine learning applications 1,2,3,4,5,6 . However, current approaches to efficiently train these models are limited by in silico emulation on digital computers. Here we develop a method called fully forward mode (FFM) learning, which implements the compute-intensive training process on the physical system. The majority of the machine learning operations are thus efficiently conducted in parallel on site, alleviating numerical modelling constraints. In free-space and integrated photonics, we experimentally demonstrate optical systems with state-of-the-art performances for a given network size. FFM learning shows training the deepest optical neural networks with millions of parameters achieves accuracy equivalent to the ideal model. It supports all-optical focusing through scattering media with a resolution of the diffraction limit; it can also image in parallel the objects hidden outside the direct line of sight at over a kilohertz frame rate and can conduct all-optical processing with light intensity as weak as subphoton per pixel (5.40 × 10 18 - operations-per-second-per-watt energy efficiency) at room temperature. Furthermore, we prove that FFM learning can automatically search non-Hermitian exceptional points without an analytical model. FFM learning not only facilitates orders-of-magnitude-faster learning processes, but can also advance applied and theoretical fields such as deep neural networks, ultrasensitive perception and topological photonics.

Fast adaptation of multi-task meta-learning for optical performance monitoring

Single-shot Autofocusing in Light Sheet Fluorescence Microscopy with Multiplexed Structured Illumination and Deep Learning

Intelligent optical performance monitor using multi-task learning based artificial neural network

Joint Optical Performance Monitoring and Modulation Format/Bit-Rate Identification by CNN-Based Multi-Task Learning

Transfer learning simplified multi-task deep neural network for PDM-64QAM optical performance monitoring

Feature Fusion-Based Multi-Task ConvNet for Simultaneous Optical Performance Monitoring and Bit-Rate/Modulation Format Identification

New Environment Adaptation with Few Shots for OFDM Receiver and mmWave Beamforming

Fast adaptive parallel computational ghost imaging based on meta learning

Fully forward mode training for optical neural networks

Multichannel meta-imagers for accelerating machine vision

Fast Adaptation for Deep Learning-based Wireless Communications

Fast Self-Learning Modulation Recognition Method for Smart Underwater Optical Communication Systems

Joint Fiber Nonlinear Noise Estimation, OSNR Estimation and Modulation Format Identification Based on Asynchronous Complex Histograms and Deep Learning for Digital Coherent Receivers

3FM: Multi-modal Meta-learning for Federated Tasks

Boosting Few-Shot Learning with Task-Adaptive Multi-level Mixed Supervision.

Faster Meta Update Strategy for Noise-Robust Deep Learning.

Multi-Learner Based Deep Meta-Learning for Few-Shot Medical Image Classification

Supervised Fine-Tuning Achieve Rapid Task Adaption Via Alternating Attention Head Activation Patterns

Intelligent Multi-channel Meta-imagers for Accelerating Machine Vision

Leveraging Multiplexed Metasurfaces for Multi-Task Learning with All-Optical Diffractive Processors

Transfer learning assisted convolutional neural networks for modulation format recognition in few-mode fibers