Fig. 1

Workflow of the magnetization reversal process analysis using the eX-GL model. The eX-GL model can address the complex inhomogeneity of the structure and establish a connection to the functions from a data science perspective. PH was employed to extract the features of the experimental magnetic domain structures, and dimensionality reduction was performed using highly interpretable machine learning. The energy landscape was drawn in the data space using these physically meaningful features. An energy landscape analysis was performed based on the relationship between the energy and features, which was used to visualize the origin of the macroscopic functions in the original microscopic structure.