An Analytical Approach to Single Node Delay-Coupled Reservoir Computing

摘要: Reservoir computing has been successfully applied in difficult time series prediction tasks by injecting an input signal into a spatially extended reservoir of nonlinear subunits to perform history-dependent computation. Recently, the network was replaced single node, delay-coupled itself. Instead spatial topology, are arrayed along one delay span system. As result, exists only implicitly differential equation, numerical solving which is costly. We derive here approximate analytical equations for underlying system explicitly. The approximation represents accurately and yields comparable performance benchmark tasks, while reducing computational costs several orders magnitude. This important implications with respect electronic realizations opens up new possibilities optimization theoretical investigation.