Optical coprocessor based on spontaneous Brillouin scattering

TL;DR

This paper proposes an optical coprocessor utilizing spontaneous Brillouin scattering in ring resonators to perform matrix-vector multiplication efficiently for neural network applications.

Contribution

It introduces a novel optical architecture using Brillouin interactions for high-speed, low-energy matrix computations in neural networks.

Findings

Demonstrates matrix-vector multiplication using ring resonators with Brillouin scattering.

Shows the system enables parallel computations in the frequency domain.

Provides a framework for implementing optical neural network coprocessors.

Abstract

Analog coprocessors for neural networks are an intensively developing field. They provide approximate results of computations for relatively low energy cost and at high speed. We show that a set of ring resonators with Brillouin interaction between photons and phonons, being coupled to a waveguide, can be used to implement matrix-vector multiplication. The input vector is formed by occupancies of the anti-Stokes optical modes pumped via spontaneous Brillouin scattering, i.e, scattering on thermal phonons. Brillouin scattering rates and coupling constants between ring resonators and the waveguide form the matrix. The system allows for parallel computations in frequency band.