Reservoir computing using high order synchronization of coupled oscillators

A A Velichko,S D Khanin,A G Rikkiev,D V Ryabokon,A V Sidorenko

doi:10.1088/1757-899x/862/5/052062

Abstract

We propose a concept for reservoir computing on oscillators using the high-order synchronization effect. The reservoir output is presented in the form of oscillator synchronization metrics: fractional high-order synchronization value and synchronization efficiency, expressed as a percentage. Using two coupled relaxation oscillators built on VO2 switches, we created an oscillator reservoir that allows simulating the XOR operation. The reservoir can operate as with static input data (power currents, coupling forces), as with dynamic data in the form of spike sequences. Having a small number of oscillators and significant non-linearity, the reservoir expresses a wide range of dynamic states. The proposed computing concept can be implemented on oscillators of diverse nature.

Highlights

Neural networks are effective tools for solving problems of pattern recognition and artificial intelligence [1]
A typical architecture for reservoir calculations consists of the following parts (Figure 1): 1. An input single-layer neural network has a matrix of weights WI and reflects the input data to the internal nodes of the reservoir
Earlier [18], we proposed a family of metrics consisting of two parameters — the high-order synchronization value SHR and the synchronization efficiency value μ

Summary

Introduction

Neural networks are effective tools for solving problems of pattern recognition and artificial intelligence [1]. An input single-layer neural network has a matrix of weights WI and reflects the input data to the internal nodes of the reservoir. 3. The readout layer is a trained neural network with a matrix of weights WR, and it converts the output of the reservoir into the resulting response of the entire system. The main task of the reservoir is to transfer the input data into a higher dimensional space that the output neural network can classify the result more accurately. Variety of existing oscillator systems determines the need of developing a new concept of reservoir computing, proposed in this study, which can be applied to oscillators of any nature. We propose a new concept of the oscillator reservoir, which operation is based on the high-order synchronization effect and special synchronization metrics. To illustrate the operation of a reservoir, we implement the XOR operation

High order synchronization of coupled oscillators reservoir

Discussion and conclusion