发表机构
Fujitsu Research, Fujitsu Limited; School of Chemical Engineering, Aalto University（富士通研究所; 阿尔托大学化学工程学院）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
针对机器学习原子间势训练中的数据选择，提出基于k-medoids聚类和最优匹配的结构相似性评估方法，兼顾表征保真与计算效率，较基线快约50倍，提升力预测的数据效率与稳定性。
AI 中文摘要
机器学习原子间势（MLIPs）是加速模拟驱动材料设计的关键组成部分。数据高效的MLIP训练依赖于数据选择策略，这些策略在限制计算成本高昂的第一性原理计算的同时，最大化结构多样性。此类策略中的一个关键挑战是评估结构相似性，这涉及在保留单个原子环境的信息与降低计算成本之间进行权衡。在此，我们提出了能够同时实现表征保真度和计算效率的相似性评估方法。我们的方法使用由k-medoids聚类识别出的一小组特征原子环境来表示每个结构，并通过这些代表或其分布之间的最优匹配来计算成对相似性。分子基准测试表明，我们的方法比基线方法快约50倍，同时也能更清晰地区分具有不同化学成分的结构。基于相似性的数据选择基准测试表明，我们的方法提高了MLIPs中力预测的数据效率和稳定性。
英文摘要
Machine learning interatomic potentials (MLIPs) are essential components for accelerating simulation-driven materials design. Data-efficient MLIP training relies on data-selection strategies that maximize structural diversity while limiting computationally expensive first-principles calculations. A key challenge in such strategies is evaluating structural similarity, which involves a trade-off between retaining information on individual atomic environments and reducing computational cost. Here, we propose similarity evaluation methods that achieve both representational fidelity and computational efficiency. Our method represents each structure using a small set of characteristic atomic environments identified by k-medoids clustering and computes pairwise similarity through optimal matching between these representatives or their distributions. Molecular benchmarks demonstrate that our method is approximately 50 times faster than the baseline method while also more clearly distinguishing structures with different chemical compositions. Similarity-based data-selection benchmarks demonstrate that our methods improve the data efficiency and stability of force prediction in MLIPs.
Comments31 pages, 12 figures
