Calculating Data Importance using Mutual Information for Engineering Design

Abstract

As engineers examine larger coupled systems, computational complexity, available resources and the lack of expert intuition create a need to understand the importance of each link of data passed through an analysis. A better understanding and an automated calculation of this data importance would enable an advance of the art for automated decomposition and optimization methods. Larger coupled problems may, for instance, expand beyond an expert's experience in manual decomposition. By automatically discovering the importance and interrelated structure of a problem, low ranked data links might be temporarily separated to decompose a problem into sub-problems. A better understanding of the larger problem may also allow for organizational optimization around the coupled sub-problems discovered by this method, that is theoretically grounded in Information Theory.