Motivation and Objectives
This project aims to develop new graph-based approaches to materials design through scalable interactive data mining and visualization techniques. The main idea is to represent a material as a network of grains — two grains are connected in the graph if they are neighbors in the material. This material-to-graph transformation enables many new exciting possibilities (see figure below). For example, the graph-based representation is potentially a much more compact representation of the material microstructures that both models and quantifies their salient features, allowing multiple materials to be compared (i.e., comparing their graph-based representations) using state-of-the-art graph similarity and matching techniques. Finding important, interesting grain structures (e.g., those that correlate with desirable structural properties) becomes finding their corresponding subgraphs of grains, which may be solved through graph querying techniques.
- Investigating how to best transform materials into their corresponding graph-based representations (e.g., what will each node in the graph represents, the semantics of a edge, etc.)
- Developing fast algorithms for graph matching (e.g., comparing materials) and querying (e.g., finding interesting grain structures)
- Developing interactive visualization that enables material scientists to easily visualize the materials and their graph-based representation, and to perform querying and matching.
The above inter-related research thrusts will create new kinds of interactive, scalable, and visual tools that will accelerate materials understanding and catalyze discovery of new materials.