Reservoir Computing Approaches for Representation and Classification of Multivariate Time Series

Classification of multivariate time series (MTS) has been tackled with a large variety methodologies and applied to wide range scenarios. Reservoir computing (RC) provides efficient tools generate vectorial, fixed-size representation the MTS that can be further processed by standard classifiers. Despite their unrivaled training speed, classifiers based on RC architecture fail achieve same accuracy fully trainable neural networks. In this article, we introduce reservoir model space, an unsupervised approach learn vectorial representations MTS. Each is encoded within parameters linear trained predict low-dimensional embedding dynamics. Compared other methods, our space yields better attains comparable computational performance due intermediate dimensionality reduction procedure. As second contribution, propose modular framework for classification, associated open-source Python library. The different modules seamlessly implement advanced architectures. architectures are compared classifiers, including deep learning models kernels. Results obtained benchmark real-world data sets show dramatically faster and, when implemented using proposed representation, also superior classification accuracy.