Multi-reservoir Echo State Network with Sparse Bayesian Learning

摘要

A multi-reservoir Echo State Network based on the Sparse Bayesian method (MrBESN) is proposed in this paper. When multivariate time series are predicted with single reservoir ESN model, the dimensions of phase-space reconstruction can be only selected a single value, which can not portray respectively the dynamic feature of complex system. To some extent, that limits the freedom degree of the prediction model and has bad effect on the predicted result. MrBESN will expand the simple input into high-dimesional feature vector and provide the automatic estimation of the hyper-parameters with Sparse Bayesian. A simulation example, that is a set of real world time series, is used to demonstrate the validity of the proposed method.