Photonic Delay Systems as Machine Learning Implementations

TL;DR

This paper demonstrates that photonic delay systems can be effectively optimized using gradient descent and backpropagation through time, significantly enhancing their performance for machine learning tasks beyond traditional reservoir computing.

Contribution

It introduces the use of gradient descent with backpropagation through time to optimize input encoding in photonic delay systems, extending their applicability in machine learning.

Findings

Optimized input encodings improve system performance.

Physical experiments confirm real-world effectiveness.

Gradient-based optimization outperforms traditional reservoir computing.

Abstract

Nonlinear photonic delay systems present interesting implementation platforms for machine learning models. They can be extremely fast, offer great degrees of parallelism and potentially consume far less power than digital processors. So far they have been successfully employed for signal processing using the Reservoir Computing paradigm. In this paper we show that their range of applicability can be greatly extended if we use gradient descent with backpropagation through time on a model of the system to optimize the input encoding of such systems. We perform physical experiments that demonstrate that the obtained input encodings work well in reality, and we show that optimized systems perform significantly better than the common Reservoir Computing approach. The results presented here demonstrate that common gradient descent techniques from machine learning may well be applicable on physical neuro-inspired analog computers.