Sample-efficient active learning for materials informatics using integrated posterior variance

Ramsey Issa

doi:10.1016/j.commatsci.2026.114551

Sample-efficient active learning for materials informatics using integrated posterior variance

2026-02-09

0

OA

AI

DOI：10.1016/j.commatsci.2026.114551

Original

Save

View PDF

Abstract

En 中文

• Active learning - Compares Integrated Posterior Variance (IPV) against random sampling, point-wise uncertainty sampling, and query-by-committee. • Multi-dataset evaluation - Benchmarked on three diverse datasets; AutoAM, Thermoelectric, and NMR. • Performance gain - IPV consistently selects candidates that reduce prediction error with fewer labeled samples than other selection strategies. • Limitations - (1) Computational overhead grows with dataset size; (2) Effectiveness declines in high-dimensional spaces where distance metrics degrade. • Practical impact - Enables strategic experiment selection that cuts required experiments for materials discovery workflows.

Keywords:

Machine learning

Artificial intelligence

Active learning

Materials science

Uncertainty

Sample-efficient active learning for materials informatics using integrated posterior variance

Abstract

AI Summary

Journal

Organization