A strategic roadmap for an atomistic machine-learning ecosystem

2026-09-30

AI Summary

This work addresses the challenges of modeling choices, software interoperability, and hardware acceleration in deploying machine learning ensembles for atomistic simulations. Building upon outcomes from a CECAM workshop, it presents the first systematic integration of physical modeling frameworks with modern deep learning techniques. By synergizing first-principles calculations, molecular dynamics, and statistical sampling with contemporary hardware accelerators, this project establishes interdisciplinary collaborative mechanisms to balance accuracy, efficiency, and scalability. The primary deliverable is a strategic roadmap encompassing both long-term objectives and concrete actions, designed to drive the co-evolution of algorithms, models, and software-hardware infrastructure. Ultimately, this initiative fosters unified community planning and efficient collaboration, setting foundational standards for establishing a robust and sustainable machine learning ecosystem at the atomic scale.

Abstract

Data-driven machine learning (ML) techniques have become an essential tool in many domains of science. Their application to atomistic simulations of matter is particularly widespread and impactful. This success is due largely to the existence of a well-developed and established physics-based modeling framework, ranging from first-principles electronic-structure calculations to molecular dynamics and statistical sampling, into which ML was integrated naturally to reshape long-standing trade-offs between accuracy, efficiency, and scale. Nevertheless, this integration raises both conceptual and practical challenges, from choosing between data-centric and physics-based modeling approaches to adapting established software stacks to modern hardware accelerators and ML libraries. As the field evolves rapidly, fueled in part by widespread enthusiasm but also by tangible impact, it seems appropriate to take a moment to consider the current state of the art and open challenges, and reflect on what can be done to better coordinate efforts across the community. With this goal in mind, several members of this community met in Lausanne in January 2026 at CECAM to discuss algorithms, models, software and hardware infrastructure, and the most promising scientific applications that have become possible thanks to the use of artificial intelligence in atomic-scale simulations. This strategic roadmap paper summarizes the outcomes of these discussions, suggesting some long-term goals, and some concrete actions, to establish a healthy, sustainable and impactful atomistic ML ecosystem.

Problem

• atomistic machine learning
• molecular simulations
• ecosystem roadmap
• physics-based modeling
• community coordination

Innovation

• Atomistic Machine Learning
• Strategic Roadmap
• Physics-based Modeling
• Molecular Dynamics
• Hardware Accelerators