Benchmarking Active Learning Strategies for Materials Optimization and Discovery. Issue 1 (9th July 2022)

• Record Type: Journal Article
• Authors: Wang, Alex; Liang, Haotong; McDannald, Austin; Takeuchi, Ichiro; Kusne, A Gilad
• Is Part Of: Oxford open materials science. Volume 2:Issue 1(2022)
• Publication Date: 2022-07-09
• Subjects: Reference data -- machine learning -- phase map -- materials optimization -- benchmark -- acquisition function; Materials science -- Periodicals; 620.11
• DOI: 10.1093/oxfmat/itac006
• Languages: English
• ISSNs: 2633-6979

Abstract

Abstract: Objectives: Autonomous physical science is revolutionizing materials science. In these systems, machine learning controls experiment design, execution, and analysis in a closed loop. Active learning, the machine learning field of optimal experiment design, selects each subsequent experiment to maximize knowledge toward the user goal. Autonomous system performance can be further improved with implementation of scientific machine learning, also known as inductive bias-engineered artificial intelligence, which folds prior knowledge of physical laws (e.g., Gibbs phase rule) into the algorithm. As the number, diversity, and uses for active learning strategies grow, there is an associated growing necessity for real-world reference datasets to benchmark strategies. We present a reference dataset and demonstrate its use to benchmark active learning strategies in the form of various acquisition functions. Methods: Active learning strategies are used to rapidly identify materials with optimal physical properties within a compositional phase diagram mapping a ternary materials system. The data is from an actual Fe-Co-Ni thin-film library and includes previously acquired experimental data for materials compositions, X-ray diffraction patterns, and two functional properties of magnetic coercivity and the Kerr rotation. Results: Popular active learning methods along with a recent scientific active learning method are benchmarked for their materials optimization performance. Conclusion: Among the acquisition functions benchmarked, Expected Improvement demonstrated the best overall performance. We discuss the relationship between algorithm performance, materials search space complexity, and the incorporation of prior knowledge, and we encourage benchmarking more and novel active learning schemes.