Visualising Temporal Data Using Reservoir Computing

摘要

We create an artificial neural network which is a version of echo state machines, ESNs. ESNs are recurrent neural networks but unlike most recurrent networks, they come with an efficient training method. We adapt this method using ideas from the neu-roscale algorithm so that the network is optimal for projecting multivariate time series data onto a low dimensional manifold so that the structure in the time series can be identified by eye. We illustrate the resulting projections on real and artificial data. Finally we compare visualisation by the technique described herein with visualisation with various standard techniques and demonstrate that the method described in this paper is better.