Revisiting Computational Thermodynamics through Machine Learning of High-Dimensional Data

This work presents a new perspective on alloy thermodynamics computation using data-driven analysis and machine learning for the design and discovery of materials. The focus is on an integrated machine-learning framework, coupling different genres of supervised and unsupervised informatics techniques and bridging two distinct viewpoints: continuum representations based on solid solution thermodynamics, and discrete high-dimensional elemental descriptions.