AIPHAD, an active learning web application for visual understanding of phase diagrams

This study introduces AIPHAD (Artificial Intelligence technique for PHAse Diagram), an open-source web application using active learning to expedite the investigation and visualization of phase diagrams. AIPHAD employs the PDC (Phase Diagram Construction) algorithm, leveraging uncertainty sampling and machine learning techniques (Label Propagation and Label Spreading) to suggest informative experiments. Its efficacy is demonstrated through the study of the Fe-Ti-Sn ternary system, efficiently identifying the Heusler phase. The integration of machine learning with traditional materials science accelerates materials exploration.

Ryo Tamura, Haruhiko Morito, Guillaume Deffrennes, Masanobu Naito, Yoshitaro Nose, Taichi Abe, Kei Terayama