Iterative Pretraining Framework for Interatomic Potentials

2025-07-26

AI Summary

Machine learning interatomic potentials (MLIPs) suffer from heavy reliance on large-scale labeled datasets and poor adaptability of general-purpose pre-trained models to downstream tasks. To address these challenges, this paper proposes the Iterative Pre-training and Inference Pipeline (IPIP), a novel framework integrating iterative optimization, an explicit forgetting mechanism, and task-adaptive fine-tuning within a lightweight architecture. The forgetting mechanism mitigates overfitting and local optima traps, thereby enhancing both accuracy and efficiency in molecular dynamics simulations. Evaluated on the Mo–S–O system, IPIP achieves over 80% reduction in energy and force prediction errors compared to conventional force fields and state-of-the-art pre-trained MLIPs, while accelerating inference by a factor of four. These improvements significantly strengthen domain-specific generalization capability without compromising computational efficiency.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Natural Language Processing: Learning & Optimization for NLP
• Search and Optimization: Learning to Search

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Web Mining and Content Analysis: Large pretrained models with web data
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Machine learning interatomic potentials (MLIPs) enable efficient molecular dynamics (MD) simulations with ab initio accuracy and have been applied across various domains in physical science. However, their performance often relies on large-scale labeled training data. While existing pretraining strategies can improve model performance, they often suffer from a mismatch between the objectives of pretraining and downstream tasks or rely on extensive labeled datasets and increasingly complex architectures to achieve broad generalization. To address these challenges, we propose Iterative Pretraining for Interatomic Potentials (IPIP), a framework designed to iteratively improve the predictive performance of MLIP models. IPIP incorporates a forgetting mechanism to prevent iterative training from converging to suboptimal local minima. Unlike general-purpose foundation models, which frequently underperform on specialized tasks due to a trade-off between generality and system-specific accuracy, IPIP achieves higher accuracy and efficiency using lightweight architectures. Compared to general-purpose force fields, this approach achieves over 80% reduction in prediction error and up to 4x speedup in the challenging Mo-S-O system, enabling fast and accurate simulations.

Problem

• Improves MLIP performance without large labeled datasets
• Reduces mismatch between pretraining and downstream tasks
• Enhances accuracy and efficiency in molecular dynamics simulations

Innovation

• Iterative Pretraining Framework for MLIPs
• Forgetting mechanism avoids local minima
• Lightweight architectures enhance accuracy efficiency