Deep Reservoir Computing using Echo State Networks and Liquid State Machine

Reservoir Computing is one of the most recent Machine Learning (ML) algorithm used for time-series prediction. Its main advantage is that it uses a fixed set of weights for the internal network. This paper was focused on Deep Reservoir Computing architectures, starting from simple Echo State Networks (ESN) and Liquid State Machine (LSM) models. The performance of such hybrid architectures was then compared with simple ESN. The final goal was to find the optimal configuration which obtains the best results in terms of training time, prediction time and Root Mean Square Error (RMSE).