Visual Analysis of Flow Features Using Information Theory

Unsteady scientific data is one of the top challenges in visualization, as a huge amount of information has to be displayed. $epsilon$-Machines are an information-theoretic concept that compress the dynamics in the data set to a finite state machine, where nodes represent local flow patterns and edges transitions between them. In this paper, we propose several enhancements to the fundamental $epsilon$-machine representation to identify interesting time intervals, analyze the evolution of unusual local dynamics and track features over time. Automatically abstracting information from the original data, as provided by these tasks, is a first step towards knowledge-assisted visualization. Successive findings in the analysis process can be used to provide subsequent users with knowledge gained in earlier research, resulting in a knowledge-assisted system for the analysis of unsteady flow features based on information theory.