This project will develop a new class of techniques suitable for analyzing the massive, complex datasets currently generated with high-performance computing resources. Specifically, the major challenge with these data is not just that they are massive in scale, but that they are multifaceted as well. Multifaceted data represent multiple values of interest simultaneously. For example, a typical, Earth system model might seek to understand the connections between (a) arctic sea ice, (b) the currents and temperature of the ocean, (c) surface temperature models, (d) atmospheric behavior that affects the surface, ocean, and ice, and (e) solar radiation's effect on the atmosphere. These values may be data from multiple sources, may encode variability or uncertainty across parameters, or may be the result of multiple physical models being computed simultaneously. Analyzing multifaceted data presents new challenges in that we seek not just to understand each facet of data, but rather we also want to understand the interactions and relationships among facets. The proposed solution develops a new field of analysis techniques called topological analytics, which couples machine learning (techniques for building statistical relationships) with topological analysis (techniques for analyzing the features of a single facet of data) to develop new analyses for multifaceted data.
