Abstract

The study of deep recurrent neural networks (RNNs) and, in particular, ofdeep Reservoir Computing (RC) is gaining an increasing research attention inthe neural networks community. The recently introduced Deep Echo State Network(DeepESN) model opened the way to an extremely efficient approach for designingdeep neural networks for temporal data. At the same time, the study of DeepESNsallowed to shed light on the intrinsic properties of state dynamics developedby hierarchical compositions of recurrent layers, i.e. on the bias of depth inRNNs architectural design. In this paper, we summarize the advancements in thedevelopment, analysis and applications of DeepESNs.

 

Quick Read (beta)