🤖 AI Summary
This work proposes an active learning framework to reduce the cost of first-principles calculations required for training machine learning interatomic potentials (MLIPs). By integrating compositional and property-based descriptors, the approach employs neural network ensembles with Query-by-Committee to quantify predictive uncertainty. The study systematically evaluates sampling strategies—including diversity-based methods (k-means and farthest point sampling), uncertainty-based selection, and hybrid approaches—across multiple material systems such as carbon, silicon, iron, and titanium oxides. For the first time in multi-material settings, diversity sampling is shown to significantly outperform alternatives, achieving target accuracy with 5–13% fewer labeled samples on average; notably, it yields a 10.9% performance gain for titanium oxides (p = 0.008). The entire workflow completes within four hours on a standard Google Colab instance with 8 GB memory.
📝 Abstract
Efficient materials discovery requires reducing costly first-principles calculations for training machine-learned interatomic potentials (MLIPs). We develop an active learning (AL) framework that iteratively selects informative structures from the Materials Project and Open Quantum Materials Database (OQMD) using compositional and property-based descriptors with a neural network ensemble model. Query-by-Committee enables real-time uncertainty quantification. We compare four strategies: random sampling (baseline), uncertainty-based sampling, diversity-based sampling (k-means clustering with farthest-point refinement), and a hybrid approach. Experiments across four material systems (C, Si, Fe, and TiO2) with 5 random seeds demonstrate that diversity sampling achieves competitive or superior performance, with 10.9% improvement on TiO2. Our approach achieves equivalent accuracy with 5-13% fewer labeled samples than random baselines. The complete pipeline executes on Google Colab in under 4 hours per system using less than 8 GB RAM, democratizing MLIP development for resource-limited researchers. Open-source code and configurations are available on GitHub. This multi-system evaluation provides practical guidelines for data-efficient MLIP training and highlights integration with symmetry-aware architectures as a promising future direction.