Self-contained implementation of Echo State Networks.
Hyperparameter optimization is carried out via Bayesian Optimization using Recycle Validation and K-fold cross Validation as described in Racca and Magri (https://doi.org/10.1016/j.neunet.2021.05.004).
Sparse matrix implementation provides fast training and low memory requirements.
As an example, the networks are used to time-accurately predict the chaotic Lorenz system.