Data Curation for Machine Learning Interatomic Potentials by Determinantal Point Processes
This work addresses the problem of efficient data curation for researchers developing machine learning interatomic potentials, offering an incremental improvement over existing methods.
The paper tackles the computational bottleneck in generating training datasets for machine learning interatomic potentials by applying determinantal point processes (DPPs) to select informative subsets of atomic configurations for labeling with quantum mechanical energies and forces. In experiments with hafnium oxide data, DPPs are shown to be competitive with existing approaches, leading to improved accuracy and robustness in machine learning representations of molecular systems.
The development of machine learning interatomic potentials faces a critical computational bottleneck with the generation and labeling of useful training datasets. We present a novel application of determinantal point processes (DPPs) to the task of selecting informative subsets of atomic configurations to label with reference energies and forces from costly quantum mechanical methods. Through experiments with hafnium oxide data, we show that DPPs are competitive with existing approaches to constructing compact but diverse training sets by utilizing kernels of molecular descriptors, leading to improved accuracy and robustness in machine learning representations of molecular systems. Our work identifies promising directions to employ DPPs for unsupervised training data curation with heterogeneous or multimodal data, or in online active learning schemes for iterative data augmentation during molecular dynamics simulation.