Implementation of rock-paper-scissors judgment systems with a Ag2S reservoir

There is a growing demand for physical reservoirs that operate with low power consumption and low computational cost. We have conducted researches on the basic properties of Ag2S reservoirs, which are a type of a physical reservoir. However, little research has been conducted on their applications. In this study, as a first step toward the practical application of Ag2S reservoirs, we implemented two types of rock-paper-scissors judgment systems using Ag2S reservoirs. In these experiments, we were able to demonstrate fast learning in the reservoir by comparing the results with methods using a single-layer perceptron (SLP) and a convolutional neural network (CNN). In addition we could obtain a maximum accuracy rate of about 98%.