Audio Classification with Skyrmion Reservoirs

TL;DR

This paper introduces a skyrmion-based reservoir computing system capable of high-accuracy spoken digit classification at nanosecond speeds, highlighting its potential for energy-efficient, in-materio machine learning applications.

Contribution

It presents the first high-performance skyrmion mixture reservoir for reservoir computing, achieving record accuracy and speed in spoken digit recognition tasks.

Findings

97.4% overall accuracy on spoken digit classification

Less than 1% word error rate achieved

Operates at nanosecond timescale

Abstract

Physical reservoir computing is a computational paradigm that enables spatio-temporal pattern recognition to be performed directly in matter. The use of physical matter leads the way towards energy-efficient devices capable of solving machine learning problems without having to build a system of millions of interconnected neurons. We propose a high performance "skyrmion mixture reservoir" that implements the reservoir computing model with multi-dimensional inputs. We show that our implementation solves spoken digit classification tasks at the nanosecond timescale, with an overall model accuracy of 97.4% and a less that 1% word error rate; the best performance ever reported for in-materio reservoir computers. Due to the quality of the results and the low power properties of magnetic texture reservoirs, we argue that skyrmion fabrics are a compelling candidate for reservoir computing.