Training Large-Scale Optical Neural Networks with Two-Pass Forward Propagation

TL;DR

This paper introduces Two-Pass Forward Propagation, a novel training method for Optical Neural Networks that improves efficiency, scalability, and energy use by avoiding nonlinear functions and enabling optical convolutional neural networks.

Contribution

The paper presents a new training approach for ONNs that enhances speed and scalability, and proposes a simple neural network implementation for optical CNNs.

Findings

Significant improvements in training speed and energy efficiency.

Successful implementation of optical convolutional neural networks.

Theoretical and numerical validation of the proposed methods.

Abstract

This paper addresses the limitations in Optical Neural Networks (ONNs) related to training efficiency, nonlinear function implementation, and large input data processing. We introduce Two-Pass Forward Propagation, a novel training method that avoids specific nonlinear activation functions by modulating and re-entering error with random noise. Additionally, we propose a new way to implement convolutional neural networks using simple neural networks in integrated optical systems. Theoretical foundations and numerical results demonstrate significant improvements in training speed, energy efficiency, and scalability, advancing the potential of optical computing for complex data tasks.