In the pursuit of designing complex materials with desired properties, un- derstanding their design parameter space is crucial. However, this space’s convolution often hinders comprehension of...Show moreIn the pursuit of designing complex materials with desired properties, un- derstanding their design parameter space is crucial. However, this space’s convolution often hinders comprehension of complex materials’ responses as a function of their design parameters. Machine Learning has recently emerged as a promising tool for capturing patterns in complex design spaces, although this performance often comes at the cost of interpretabil- ity. This thesis aims to explore the design parameter space of interact- ing hysterons using interpretable Machine Learning, specifically Decision Tree inspired methods. Despite the complexity of the design parameter space of even small systems of interacting hysterons, interpretable Ma- chine Learning can classify coarse-grained properties of the system effec- tively. Introducing the Support Vector Classifier (SVC) inspired Decision Tree, we achieve almost perfect isolation of these properties. This model preserves interpretability while effectively probing the statistical structure of design parameter space of systems of interacting hysterons.Show less
