A statistical approach to infer connectivity in complex networks using the mutual information rate

A statistical approach to infer connectivity in complex networks using the mutual information rate #

Huseyin Yildirim, Chris Antonopoulos

15:10 Monday in 4Q07.

Part of the Neurons and networks session.

Abstract #

The last few decades have seen the development of complex network science, which is trying to find ways to model natural phenomena in a wide range of disciplines, from Biology to Social Sciences to Neurosciences, to name a few. Among these developments is the Mutual Information Rate (MIR), that is defined as the amount of information per unit of time passing from one system component (node) to another in a network. MIR can be used to infer connectivity in complex networks using time-series data [1]. One of the main challenges in using MIR is to define appropriate thresholds to infer connectivity successfully [1]. Recently, the author in [2] proposed the use of statistical hypothesis tests and MIR to tackle this challenge. It was shown that one can infer network connectivity by comparing the MIR values of time series statistically from the complex network with MIR values of appropriately generated surrogate datasets based on the source of connectivity. These might be the correlation between magnitudes, phase synchronisation, etc. This talk presents recent findings on the use of statistical tests and MIR to infer network connectivity from recorded time series.

References:

  1. Bianco-Martinez, E., Rubido, N., Antonopoulos, C.G. and Baptista, M.S., 2016. Successful network inference from time-series data using mutual information rate. Chaos: An Interdisciplinary Journal of Nonlinear Science, 26(4), p.043102.

  2. Antonopoulos, C.G., 2023. Network inference combining mutual information rate and statistical tests. Communications in Nonlinear Science and Numerical Simulation, 116, p.106896