Inferring directional interactions in collective dynamics

Establishing whether directional interactions are present between coupled systems is a fundamental problem in dynamical systems theory. New tools such as intrinsic mutual information have been recently introduced to measure information flow between two systems. However, its use in the inference of directional interactions is yet to be determined.

In their new manuscript, our own Professor Maurizio Porfiri with Professor Pietro De Lellis (University of Naples Federico II) and Manuel Ruiz Marín (Technical University of Cartagena, Murcia) provided a comparison between intrinsic mutual information and two classical metrics, time-delayed mutual information and transfer entropy. Through a tractable model of leader-follower interactions, they showed that intrinsic mutual information suffers from the same problems of time-delayed mutual information, leading to an excess of false positives compared to transfer entropy.

Read the paper “Inferring directional interactions in collective dynamics: a critique to intrinsic mutual information” in the latest issue of Journal of Physics: Complexity here.

Image credit: AdobeStock/tomertu.