EchoBay is a C++, Eigen powered library, for the training and deployment of Echo State Networks, with multiple layers and different topologies. EchoBay employs Limbo library to find the best set of hyper-parameters that maximize a score function. EchoBay is designed to work almost seamlessly on small computers, microcontrollers (tested on ESP32) and large multi-threaded systems.
Optional dependencies to read CSV data fast-cpp-csv-parser.
Full documentation is available here.
After configuration and building (see docs for details) the basic usage is:
./echobay train configuration.yml outputFolderName
./echobay compute modelFolderName
Other important contributions in this research are from Claudio Gallicchio and Alessio Micheli (Department of Computer Science, University of Pisa) and Marco D. Santambrogio (Dipartimento di Informatica, Elettronica e Bioingegneria, Politecnico di Milano)
If you use EchoBay in a scientific paper, please cite:
Cerina, Luca, Giuseppe Franco, and Marco D. Santambrogio. (2019) Lightweight autonomous bayesian optimization of Echo-State Networks., European Symposium on Artificial Neural Networks, ESANN, 2019.