[Paper Review] Data-driven active learning approaches for accelerating materials discovery

Jiaxin Chen, Tianjiao Wan+11 more|arXiv (Cornell University)|Jan 11, 2026

Machine Learning in Materials Science1 citations

TL;DR

A comprehensive review of active learning (AL) methods—traditional and deep learning-based—for improving data efficiency and accelerating materials discovery across simulation, design, optimization, and autonomous laboratories.

ABSTRACT

Materials discovery is a cornerstone of modern technological advancement, yet it remains constrained by traditional trial-and-error paradigms and the inherent bias of human intuition. Artificial intelligence (AI) has emerged as a transformative tool in materials science by effectively modeling structure-property relationships. Despite substantial efforts to enhance model expressiveness, data efficiency remains an equally critical challenge, given the limited availability of experimental and computational resources. Active learning (AL), as a data-driven machine learning paradigm, has shown great promise for discovering novel materials and enabling the efficient navigation of vast materials spaces. In this review, we follow the evolution of sampling strategy design techniques in AL, from Bayesian optimization to advanced deep learning-based strategies. We then highlight how AL enhances data efficiency across various data regimes, ranging from task-specific settings with limited data to the development of general-purpose datasets and large-scale models. We further provide a systematic overview of AL applications throughout the materials research pipeline, including computational simulation, composition and structural design, process optimization, and self-driving laboratory systems. Finally, we pinpoint key challenges and future perspectives of AL in materials discovery.

Motivation & Objective

• Motivate the need for data-efficient AI in materials discovery amid costly experiments and simulations.
• Survey classic and deep active learning approaches used to navigate large materials spaces efficiently.
• Explain how AL integrates with the materials research pipeline, including simulation, design, and self-driving laboratories.
• Identify challenges and future directions for robust, scalable AL tools in materials science.

Proposed method

• Classify AL paradigms into stream-based, pool-based, and membership query synthesis and discuss their relevance to materials research.
• Review traditional AL methods (Gaussian processes, RFs, SVMs) and core acquisition functions (EI, PI, UCB/LCB) with domain-specific features.
• Describe deep active learning (DAL) including deep kernel learning, Bayesian neural networks, Monte Carlo dropout, and multi-fidelity MOBO in DL settings.
• Discuss uncertainty-based, distribution-based, and utility-based sampling, plus hybrid strategies to balance exploration, diversity, and exploitation.
• Explain how AL is applied across the materials research pipeline and in self-driving laboratory contexts.

Experimental results

Research questions

• RQ1 How can active learning improve data efficiency in materials discovery under limited labeling resources?
• RQ2 What are the most effective AL strategies (traditional and deep) across different data regimes and material design tasks?
• RQ3 How do AL methods integrate with computational simulation, composition/structure design, process optimization, and autonomous laboratories?
• RQ4 What are the key challenges (cold-start, distribution shift, robustness) and future directions for AL in materials science?

Key findings

• AL significantly improves data efficiency and exploration of vast materials spaces compared to exhaustive evaluation.
• Bayesian optimization, uncertainty-based, distribution-based, and utility-based approaches each offer complementary strengths and can be hybridized for robust performance.
• Deep active learning extends AL to DL models via deep kernel learning, Bayesian neural networks, ensembles, and evidential DL to enable uncertainty quantification.
• AL has broad applications across computational simulation, compositional/structural design, process optimization, and self-driving lab systems.
• Challenges include robustness, scalability, and evaluation of AL tools in materials contexts.