Robust methods to detect coupling among nonlinear dynamic time series

2024/12/15 by Timothy Sauer, Sauer, Timothy, George Sugihara +1
Computer Science · Engineering · #Chaotic Dynamics (nlin.CD) #Control Systems and Identification #FOS: Physical sciences #Fault Detection and Control Systems #Neural Networks and Applications

paper · pdf · doi:10.48550/arxiv.2412.11252

openalex publication_date 2024/12/15 · openalex created_date 2024/12/18 · openalex updated_date 2026/07/28

Abstract

Two numerical methods are proposed for detection of coupling between multiple time series generated by deterministic nonlinear systems. The first detects interdependence or the existence of coupling between time series. The second ascertains directionality of coupling, or alternatively, latent coupling, the case when multiple series are driven by another, unobserved system. In either case, the driver and the recipients of the coupling may be periodic or aperiodic, and in particular may be chaotic. The only inputs to the method are two or more simultaneously recorded time series. The methods rely solely on ranking distances between states in time-delay reconstructions of the data, and for that reason tend to be robust to observational noise.

Related

• Time series analysis of coupled slow-fast neuron models: From Hurst exponent to Granger causality
• Causality studied in reconstructed state space. Examples of uni-directionally connected chaotic systems
• Investigations of Complex Systems' Dynamics Based on a Reduced Amount of Information -- Part 3: Dynamical Phenomena Indicator -- The Fastest and Most Universal Approach for Analyzing the Dynamics of Networks of Coupled Nonlinear Systems
• Information processing by a controlled coupling process
• Model-free measure of coupling from embedding principle