Learning Materials Properties from Scarce Labels and Unlabeled Crystals
2026-08-31 • Machine Learning
Machine LearningArtificial Intelligence
AI summaryⓘ
The authors created SemiMat, a benchmark to test methods that learn material properties from limited labeled and unlabeled crystal data. They also developed MatRank, a method that assigns reliability weights to uncertain labels and improves learning by considering how much different predictions agree. Their experiments show that MatRank generally produces more accurate property predictions across multiple tasks and graph-based models. The authors provide tools to understand when their approach works well and when more testing is needed. They have made their code publicly available for others to use and build upon.
semi-supervised learningmaterials property regressionpseudo-labelinggraph neural networksnormalized mean absolute error (NMAE)benchmark datasetsuncertainty weightingout-of-distribution detectionmaterial sciencemachine learning
Authors
Wentao Li, Yizhe Chen, Jiangjie Qiu, Yijun Li, Leyi Zhao, Xiaonan Wang
Abstract
Learning materials properties from scarce labels and unlabeled crystals is a central challenge for data-driven materials discovery. We present SemiMat, a controlled benchmark for semi-supervised materials property regression, and MatRank, a reliability-weighted objective for continuous pseudo-label uncertainty. SemiMat fixes labeled and unlabeled crystal inputs, graph-backbone interfaces, validation-only checkpoint selection, held-out test reporting, normalized MAE (NMAE), and method-rank summaries across six scarce-label tasks, four graph backbones, and five predefined split runs. MatRank builds pseudo-targets from labeled anchors, weights them by local reliability and weak-prediction agreement, trains weak and strong graph views consistently, and adds ranking signals so that unlabeled crystals shape both values and candidate order. Across the retained 24 backbone-task blocks, one fixed MatRank objective gives the lowest aggregate held-out test NMAE (0.896) and best average method rank (2.208). The component, OOD, and generated-pool diagnostics identify where the gain is reliable and where further screening evaluation remains necessary. Code is available at https://github.com/littlepeachs/SemiMat.