Predicting high-dimensional heterogeneous time series employing generalized local states

TL;DR

This paper introduces generalized local states (GLS) for predicting high-dimensional, heterogeneous chaotic time series, leveraging nonlinear correlation-based distances, and demonstrates improved prediction and separation capabilities over traditional local states using reservoir computing.

Contribution

The paper develops a novel GLS framework that extends local states for better prediction and separation of complex mixed chaotic systems, enhancing reservoir computing methods.

Findings

GLS improves prediction accuracy for heterogeneous systems

GLS enables separation of different chaotic systems

Prediction with GLS succeeds where LS fails

Abstract

We generalize the concept of local states (LS) for the prediction of high-dimensional, potentially mixed chaotic systems. The construction of generalized local states (GLS) relies on defining distances between time series on the basis of their (non-)linear correlations. We demonstrate the prediction capabilities of our approach based on the reservoir computing (RC) paradigm using the Kuramoto-Sivashinsky (KS), the Lorenz-96 (L96) and a combination of both systems. In the mixed system a separation of the time series belonging to the two different systems is made possible with GLS. More importantly, prediction remains possible with GLS, where the LS approach must naturally fail. Applications for the prediction of very heterogeneous time series with GLSs are briefly outlined.