A Local Approach for Information Transfer

TL;DR

This paper introduces a local method to estimate information transfer in complex systems using time series data, leveraging nearest neighbors to infer dynamical rules and probability densities, validated on simulated and real data.

Contribution

It presents a novel local approach for measuring information transfer based on nearest neighbors, improving analysis of complex systems from time series.

Findings

Method performs well on simulated data

Effective on real signals

Provides detailed local information transfer estimates

Abstract

In this work, a strategy to estimate the information transfer between the elements of a complex system, from the time series associated to the evolution of this elements, is presented. By using the nearest neighbors of each state, the local approaches of the deterministic dynamical rule generating the data and the probability density functions, both marginals and conditionals, necessaries to calculate some measures of information transfer, are estimated. The performance of the method using numerically simulated data and real signals is exposed.