Semiconductor technologies and related topics for implementation of electronic reservoir computing systems

Abstract

Reservoir computing (RC) is a unique machine learning framework based on a recurrent neural network, which is currently involved in numerous research fields. RC systems are distinguished from other machine learning systems since detailed network designs and weight adjustments are not necessary. This enables the availability of many device and material options to physically implement the system, referred to as physical RC. This review outlines the basics of RC and related issues from an implementation perspective that applies semiconductor electron device technology. A possible interpretation of RC computations is shown using a simple model, and the reservoir network is understood from the viewpoint of network theory. Physical implementation and operation issues are discussed by referring to our experimental investigation of dynamic nodes using a semiconductor tunnel diode with cubic nonlinearity.