A Critical Examination of Active Learning Workflows in Materials Science

TL;DR

This paper critically examines active learning workflows in materials science, highlighting assumptions, pitfalls, and providing guidance to improve their reliability and effectiveness in applications like interatomic potentials and self-driving labs.

Contribution

It systematically assesses key design choices in AL workflows, identifying common pitfalls and offering practical strategies for better implementation in materials science.

Findings

Identifies common pitfalls in AL workflows

Provides mitigation strategies for design assumptions

Guides practitioners in AL workflow assessment

Abstract

Active learning (AL) plays a critical role in materials science, enabling applications such as the construction of machine-learning interatomic potentials for atomistic simulations and the operation of self-driving laboratories. Despite its widespread use, the reliability and effectiveness of AL workflows depend on implicit design assumptions that are rarely examined systematically. Here, we critically assess AL workflows deployed in materials science and investigate how key design choices, such as surrogate models, sampling strategies, uncertainty quantification and evaluation metrics, relate to their performance. By identifying common pitfalls and discussing practical mitigation strategies, we provide guidance to practitioners for the efficient design, assessment, and interpretation of AL workflows in materials science.