MLMOD - Machine Learning Methods for Data-Driven Modeling in LAMMPS

Aug 2, 2026

MLMOD is a software package designed to integrate machine learning approaches and learned models into LAMMPS simulations for microscale mechanics and molecular dynamics. It supports data-driven modeling for system dynamics at larger space-time scales, interactions between components, coarse-grained features, and new quantities of interest. The package provides integration hooks for dynamics/time-step integration, interaction modeling, and quantity-of-interest computation, enabling training with general ML model classes such as neural networks and Gaussian process methods. A prototype C++/Python implementation is discussed, with source code archived and example workflows outlined.

LanguageEnglish

CategoryTechnology

FormatPDF

Pages9

Reading Time23 min

AccessFree to download

Summary

Statement of Need

Data-Driven Modeling

arXiv :2 107 . 14362v2 [ cs .LG] 21 Oct 2023 MLMOD: Machine Learning Methods for Data-Driven Modeling in LAMMPS Paul J. Atzberger 1 1 Paul J. Atzberger, Professor, University of California Santa Barbara Summary MLMOD is a software package for incorporating machine learning approaches and models into simulations of microscale mechanics and molecular dynamics in LAMMPS. Recent machine learning approaches provide promising data-driven approaches for learning representations for system behaviors from experimental data and high fidelity simulations. The package faciliates learning and using data-driven models for (i) dynamics of the system at larger spatial-temporal scales (ii) interactions between system components, (iii) features yielding coarser degrees of freedom, and (iv) features for new quantities of interest characterizing system behaviors. MLMOD provides hooks in LAMMPS for (i) modeling dynamics and time-step integration, (ii) modeling interactions, and (iii) computing quantities of interest characterizing system states. The package allows for use of machine learning methods with general model classes including Neural Networks, Gaussian Process Regression, Kernel Models, and other approaches. Here we discuss our prototype C++/Python package, aims, and example usage. The package is integrated currently with the mesocale and molecular dynamics simulation package LAMMPS and PyTorch. The source code for this initial version 1.0 of MLMOD has been archived to Zenodo (P. J. Atzberger, 2023) . For related papers, examples, updates, and additional information see [https://github.com/atzberg/mlmod](https://github.com/atzberg/mlmod) and [http://atzberger.org/](http://atzberger.org/) . Statement of Need A practical challenge in using machine learning methods for simulations is the efforts required to incorporate learned system features to augment existing models and simulation methods. Our package MLMOD aims to address this aspect of data-driven modeling by providing a general interface for incorporating ML models using standardized representations and by leveraging existing simulation frameworks such as LAMMPS (Thompson et al., 2022) . Our MLMOD package provides hooks which are triggered during key parts of simulation calculations. In this way standard machine learning frameworks can be used to train ML models, such as PyTorch (Paszke et al., 2019) and TensorFlow (Abadi et al., 2015), with the resulting models more amenable to being translated into practical simulations. The models obtained from learning can be accommodated in many forms, including Deep Neural Networks (DNNs)(Goodfellow et al., 2016), Kernel Regression Models (KRM) (Scholkopf & Smola, 2001), Gaussian Process Regression (GPR) (Rasmussen, 2004), and others (Hastie et al., 2001) . Data-Driven Modeling Recent advances in machine learning, optimization, and available computational resources are presenting new opportunities for data-driven modeling and simulation in the natural sciences and engineering. Empirical successes in deep learning suggest promising non-linear techniques for learning representations for system behaviors and other underlying features (Goodfellow et al., 2016; Hinton & Salakhutdinov, 2006) . Many previous deep learning methods have been developed for problems motivated by image analysis and natural language processing. However, scientific computations and associated dynamical systems present a unique set of challenges for developing and employing recent machine learning approaches (P. J. Atzberger, 2018; Brunton et al., 2016; Schmidt & Lipson, 2009) . In scientific and engineering applications there are often important constraints arising from physical principles required to obtain plausible models and there is a need for results to be more interpretable. In large-scale scientific computations, bottom-up modeling efforts aim to start as close as possible to first principles and perform computations to obtain insights into large

FAQs

What problem does MLMOD address for simulation-based machine learning?

MLMOD addresses the practical effort required to incorporate learned system features into existing simulation models. It provides standardized interfaces and hooks within simulation workflows to connect ML training and usage to LAMMPS computations.

Which parts of LAMMPS does MLMOD integrate with?

MLMOD provides hooks for modeling dynamics and time-step integration, modeling interactions, and computing quantities of interest that characterize system states during simulation.

What machine learning model types does MLMOD support?

MLMOD supports general model classes including neural networks, Gaussian process regression, kernel models, and other approaches, enabling learned representations and predictions to be embedded into simulations.

Aug 2, 2026

Upload by Emma Wilson