Physics- and data-driven Active Learning of neural network representations for free energy functions of materials from statistical mechanics

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Holber, Jamie, Garikipati, Krishna
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866910869386952704
author Holber, Jamie
Garikipati, Krishna
author_facet Holber, Jamie
Garikipati, Krishna
contents Accurate free energy representations are crucial for understanding phase dynamics in materials. We employ a scale-bridging approach to incorporate atomistic information into our free energy model by training a neural network on DFT-informed Monte Carlo data. To optimize sampling in the high-dimensional Monte Carlo space, we present an Active Learning framework that integrates space-filling sampling, uncertainty-based sampling, and physics-informed sampling. Additionally, our approach includes methods such as hyperparameter tuning, dynamic sampling, and novelty enforcement. These strategies can be combined to reduce MSE,either globally or in targeted regions of interest,while minimizing the number of required data points. The framework introduced here is broadly applicable to Monte Carlo sampling of a range of materials systems.
format Preprint
id arxiv_https___arxiv_org_abs_2503_07619
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Physics- and data-driven Active Learning of neural network representations for free energy functions of materials from statistical mechanics
Holber, Jamie
Garikipati, Krishna
Computational Physics
Machine Learning
Accurate free energy representations are crucial for understanding phase dynamics in materials. We employ a scale-bridging approach to incorporate atomistic information into our free energy model by training a neural network on DFT-informed Monte Carlo data. To optimize sampling in the high-dimensional Monte Carlo space, we present an Active Learning framework that integrates space-filling sampling, uncertainty-based sampling, and physics-informed sampling. Additionally, our approach includes methods such as hyperparameter tuning, dynamic sampling, and novelty enforcement. These strategies can be combined to reduce MSE,either globally or in targeted regions of interest,while minimizing the number of required data points. The framework introduced here is broadly applicable to Monte Carlo sampling of a range of materials systems.
title Physics- and data-driven Active Learning of neural network representations for free energy functions of materials from statistical mechanics
topic Computational Physics
Machine Learning
url https://arxiv.org/abs/2503.07619