Cutting Through the Noise: On-the-fly Outlier Detection for Robust Training of Machine Learning Interatomic Potentials
This provides an automated solution for training robust machine learning interatomic potentials on imperfect datasets, reducing the need for manual effort or expensive retraining cycles, which is incremental but practical for computational chemistry applications.
The paper tackles the problem of numerical noise in reference data for machine learning interatomic potentials, introducing an on-the-fly outlier detection method that automatically down-weights noisy samples during training, reducing energy errors by a factor of three on the SPICE dataset and recovering accurate physical observables like diffusion coefficients for liquid water.
The accuracy of machine learning interatomic potentials suffers from reference data that contains numerical noise. Often originating from unconverged or inconsistent electronic-structure calculations, this noise is challenging to identify. Existing mitigation strategies such as manual filtering or iterative refinement of outliers, require either substantial expert effort or multiple expensive retraining cycles, making them difficult to scale to large datasets. Here, we introduce an on-the-fly outlier detection scheme that automatically down-weights noisy samples, without requiring additional reference calculations. By tracking the loss distribution via an exponential moving average, this unsupervised method identifies outliers throughout a single training run. We show that this approach prevents overfitting and matches the performance of iterative refinement baselines with significantly reduced overhead. The method's effectiveness is demonstrated by recovering accurate physical observables for liquid water from unconverged reference data, including diffusion coefficients. Furthermore, we validate its scalability by training a foundation model for organic chemistry on the SPICE dataset, where it reduces energy errors by a factor of three. This framework provides a simple, automated solution for training robust models on imperfect datasets across dataset sizes.