This presentation will discuss the potential of using the framework of topological data analysis to guide the discovery of hidden patterns in data sets in the field of materials science. Examples are given in data derived from both computational studies as well experimental data, where we can extract new insights for materials discovery and design that hitherto had been impossible or very difficult to find; by uncovering geometric structures behind that data. Topological data analysis in materials science problems can be used to provide a means to identify groups or classifications of data that would otherwise be hidden; and uncover networks of information flow that elucidates the pathways through which heterogeneous and diverse data sets are connected. It is proposed that the use of the principles of computational homology to identify connectivity between disparate data can be a good way to integrate experimental design of high throughput experiments into focused physics based modeling of materials; hence permitting an iterative framework where experiment and computation are mutually guided.
Back to Machine Learning for Many-Particle Systems