LTAU-FF: Loss Trajectory Analysis for Uncertainty in atomistic Force Fields

Model ensembles are effective tools for estimating prediction uncertainty in deep learning atomistic force fields. However, their widespread adoption is hindered by high computational costs and overconfident error estimates. In this work, we address these challenges by leveraging distributions of pe...

Full description

Saved in:
Bibliographic Details
Main Authors: Joshua A Vita, Amit Samanta, Fei Zhou, Vincenzo Lordi
Format: Article
Language:English
Published: IOP Publishing 2025-01-01
Series:Machine Learning: Science and Technology
Subjects:
Online Access:https://doi.org/10.1088/2632-2153/adb4b9
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850192758647554048
author Joshua A Vita
Amit Samanta
Fei Zhou
Vincenzo Lordi
author_facet Joshua A Vita
Amit Samanta
Fei Zhou
Vincenzo Lordi
author_sort Joshua A Vita
collection DOAJ
description Model ensembles are effective tools for estimating prediction uncertainty in deep learning atomistic force fields. However, their widespread adoption is hindered by high computational costs and overconfident error estimates. In this work, we address these challenges by leveraging distributions of per-sample errors obtained during training and employing a distance-based similarity search in the model latent space. Our method, which we call LTAU (Loss Trajectory Analysis for Uncertainty), efficiently estimates the full probability distribution function of errors for any test point using the logged training errors, achieving speeds that are 2–3 orders of magnitudes faster than typical ensemble methods and allowing it to be used for tasks where training or evaluating multiple models would be infeasible. We apply LTAU towards estimating parametric uncertainty in atomistic force fields ( LTAU-FF ), demonstrating that it produces well-calibrated confidence intervals and predicts errors that correlate strongly with the true errors for data near the training domain. Furthermore, we show that the errors predicted by LTAU-FF can be used in practical applications for detecting out-of-domain data, tuning model performance, and predicting failure during simulations. We believe that LTAU will be a valuable tool for uncertainty quantification in atomistic force fields and is a promising method that should be further explored in other domains of machine learning.
format Article
id doaj-art-6e4e4dcc301a4e0c8b240a57ef087068
institution OA Journals
issn 2632-2153
language English
publishDate 2025-01-01
publisher IOP Publishing
record_format Article
series Machine Learning: Science and Technology
spelling doaj-art-6e4e4dcc301a4e0c8b240a57ef0870682025-08-20T02:14:26ZengIOP PublishingMachine Learning: Science and Technology2632-21532025-01-016101504810.1088/2632-2153/adb4b9LTAU-FF: Loss Trajectory Analysis for Uncertainty in atomistic Force FieldsJoshua A Vita0https://orcid.org/0000-0001-9191-055XAmit Samanta1https://orcid.org/0000-0003-3620-987XFei Zhou2https://orcid.org/0000-0001-9659-4648Vincenzo Lordi3https://orcid.org/0000-0003-2415-4656Materials Science Division Lawrence Livermore National Laboratory Livermore , Livermore, CA 94550, United States of AmericaPhysics Division Lawrence Livermore National Laboratory Livermore , Livermore, CA 94550, United States of AmericaPhysics Division Lawrence Livermore National Laboratory Livermore , Livermore, CA 94550, United States of AmericaMaterials Science Division Lawrence Livermore National Laboratory Livermore , Livermore, CA 94550, United States of AmericaModel ensembles are effective tools for estimating prediction uncertainty in deep learning atomistic force fields. However, their widespread adoption is hindered by high computational costs and overconfident error estimates. In this work, we address these challenges by leveraging distributions of per-sample errors obtained during training and employing a distance-based similarity search in the model latent space. Our method, which we call LTAU (Loss Trajectory Analysis for Uncertainty), efficiently estimates the full probability distribution function of errors for any test point using the logged training errors, achieving speeds that are 2–3 orders of magnitudes faster than typical ensemble methods and allowing it to be used for tasks where training or evaluating multiple models would be infeasible. We apply LTAU towards estimating parametric uncertainty in atomistic force fields ( LTAU-FF ), demonstrating that it produces well-calibrated confidence intervals and predicts errors that correlate strongly with the true errors for data near the training domain. Furthermore, we show that the errors predicted by LTAU-FF can be used in practical applications for detecting out-of-domain data, tuning model performance, and predicting failure during simulations. We believe that LTAU will be a valuable tool for uncertainty quantification in atomistic force fields and is a promising method that should be further explored in other domains of machine learning.https://doi.org/10.1088/2632-2153/adb4b9machine learninginteratomic potentialforce fielduncertainty quantificationensembles
spellingShingle Joshua A Vita
Amit Samanta
Fei Zhou
Vincenzo Lordi
LTAU-FF: Loss Trajectory Analysis for Uncertainty in atomistic Force Fields
Machine Learning: Science and Technology
machine learning
interatomic potential
force field
uncertainty quantification
ensembles
title LTAU-FF: Loss Trajectory Analysis for Uncertainty in atomistic Force Fields
title_full LTAU-FF: Loss Trajectory Analysis for Uncertainty in atomistic Force Fields
title_fullStr LTAU-FF: Loss Trajectory Analysis for Uncertainty in atomistic Force Fields
title_full_unstemmed LTAU-FF: Loss Trajectory Analysis for Uncertainty in atomistic Force Fields
title_short LTAU-FF: Loss Trajectory Analysis for Uncertainty in atomistic Force Fields
title_sort ltau ff loss trajectory analysis for uncertainty in atomistic force fields
topic machine learning
interatomic potential
force field
uncertainty quantification
ensembles
url https://doi.org/10.1088/2632-2153/adb4b9
work_keys_str_mv AT joshuaavita ltaufflosstrajectoryanalysisforuncertaintyinatomisticforcefields
AT amitsamanta ltaufflosstrajectoryanalysisforuncertaintyinatomisticforcefields
AT feizhou ltaufflosstrajectoryanalysisforuncertaintyinatomisticforcefields
AT vincenzolordi ltaufflosstrajectoryanalysisforuncertaintyinatomisticforcefields