Information Theoretic Causal Coordination
Negar Kiyavash · 2013
Abstract : Automated decision-making over large-scale distributed systems in the presence of uncertainty and incomplete information is a formidable task. The traditional view of information fusion and decision making over sensor networks is heavily biased by the fact that dependencies between information sources are treated only in terms of correlation. Inspired by brain's excellent job at processing information, we make a departure from this traditional view point by realizing that: human judgments about the likelihood of events and dependencies among variables are strongly influenced by the perception of cause-effect relationships. In terms of incorporating causality, we pursued formulation of information-theoretic causality metrics such as directed information. We developed a general framework for inferring causal influences in stochastic networks as well as information fusion in online recommendation systems.