Juan Restrepo, Department of Applied Mathematics, University of Colorado Boulder
Using machine learning to assess short term causal dependence and infer network links
The general problem of determining causal dependencies in an unknown time-evolving system from time series observations is of great interest in many fields. Examples include inferring neuronal connections from spiking data, deducing causal dependencies between genes from expression data, discovering long spatial range influences in climate variations, etc. Previous work has often tackled such problems by consideration of correlations, prediction impact, or information transfer metrics. In this talk I will present a new method that leverages the potential ability of machine learning to perform predictive and interpretive tasks and uses this to extract information on causal dependence. The method is tested on model complex systems consisting of networks of many interconnected dynamical units. These tests show that machine learning offers a unique and potentially highly effective approach to the general problem of causal inference.