Visualizing Information Conservation and Decomposition via the Information Matrix

Dor Tsur, Haim H. Permuter

Keywords: Data visualization, directed information, information theory, mutual information, transfer entropy

TL;DR: We propose an information theoretic visualization method termed the Information Matrix and use it to analyze and illustrate information transfer and conservation in time-series data.

Abstract: We introduce a novel framework for visualizing information conservation, decomposition and transfer in time-series data, termed the Information Matrix ($\bm{I}^{XY}$). Our approach, grounded in information theory, focuses on mutual information (MI), directed information (DI), and transfer entropy (TE) to analyze sequential data. This framework not only offers theoretical insights into information dynamics in sequential systems but also provides a simple visualization of information flow in such systems. We demonstrate the utility of the Information Matrix to the analysis of sequential real world data.

Submission Number: 70