A Machine Learning Tool for Materials Informatics

In response to the increasing demand for the highly efficient design of materials, materials informatics has been proposed for using data and computational sciences to extract data features that provide insight into how properties track with microstructure variables. However, the general metrics of microstructural features often ignore the complexities of the microstructure geometry for many properties of interest. An independently developed machine learning tool called shiny materials genome integration system for phase and property analysis (ShinyMIPHA), which is designed with either standalone software or cloud system based on an R programing package of "Shiny", is introduced. ShinyMIPHA provides topological microstructure analysis methods based on image processing technology by employing a two-point correlation function, persistent homology, and mean (H)-Gauss (K) curvature approaches, as well as sparse study and regression analysis methods that enable a data-driven properties-to-microstructure-to-processing inverse materials-design approach. The demo version is available at https://adachi- lab.shinyapps.io/demo/.