Towards fully integrated photonic backpropagation training and inference using on-chip nonlinear activation and gradient functions

TL;DR

This paper demonstrates the implementation of nonlinear activation functions and their gradients on a silicon photonic platform, enabling integrated photonic neural network training with high accuracy on MNIST.

Contribution

It introduces a method to realize neural nonlinearities and their gradients on-chip using silicon photonics, facilitating fully integrated photonic neural network training.

Findings

Achieved over 97% classification accuracy on MNIST.

Successfully implemented nonlinear activation functions and gradients on silicon photonics.

Enabled potential for fully integrated photonic neural network systems.

Abstract

Gradient descent-based backpropagation training is widely used in many neural network systems. However, photonic implementation of such method is not straightforward mainly since having both the nonlinear activation function and its gradient using standard integrated photonic components is challenging. Here, we demonstrate the realization of two commonly used neural nonlinear activation functions and their gradients on a silicon photonic platform. Our method leverages the nonlinear electro-optic response of a micro-disk modulator. As a proof of concept, the experimental results are incorporated into a neural network simulation platform to classify MNIST handwritten digits dataset where we classification accuracies of more than 97\% are achieved that are on par with those of ideal nonlinearities and gradients.