A Reservoir Computing Scheme for Multi-class Classification

A Reservoir Computing Scheme for Multi-class Classification

연구 분야: Infrastructure

학회: ACMSE '20: Proceedings of the 2020 ACM Southeast Conference

A recent addition to the field of neural networks is a framework for recurrent neural networks (RNNs) called reservoir computing. Reservoir computing has proven successful in predicting dynamical systems, but little attention has been given to its possible use as a classification method. Literature has been written specifically in regard to reservoir computing for binary classification, but few papers have specifically been written about multi-class classification. This article aims to provide a generic scheme for multi-class classification based on reservoir computing. The comparable performance has shown that our proposed scheme as an alternative classifier can catch up with and even outstrip the performance of some traditional classifiers such as naïve Bayesian, decision trees, random forest, and neural network over several data sets.