PulseAugur
EN
LIVE 20:31:46

New kernels from pretrained MACE potentials improve active learning for MLIPs

By PulseAugur Editorial · [2 sources] ·

Researchers have developed a new method for active learning in machine learning interatomic potentials (MLIPs) by utilizing pretrained model representations. This approach leverages the latent space of a pretrained MACE potential to generate acquisition signals, specifically a finite-width neural tangent kernel and an activation kernel. These signals effectively reduce the data required for training, showing an average reduction of 38% for energy error and 28% for force error on reactive chemistry benchmarks. AI

IMPACT This method could significantly reduce the data requirements for training MLIPs, accelerating research in materials science and chemistry.

RANK_REASON This is a research paper detailing a new method for active learning in MLIPs.

Read on arXiv cs.LG →

AI-generated summary · Google Gemini · from 2 sources. How we write summaries →

New kernels from pretrained MACE potentials improve active learning for MLIPs

COVERAGE [2]

  1. arXiv cs.LG TIER_1 English(EN) · Eszter Varga-Umbrich, Shikha Surana, Paul Duckworth, Jules Tilly, Olivier Peltre, Zachary Weller-Davies ·

    Pretrained Model Representations as Acquisition Signals for Active Learning of MLIPs

    arXiv:2605.03964v1 Announce Type: new Abstract: Training machine learning interatomic potentials (MLIPs) for reactive chemistry is often bottlenecked by the high cost of quantum chemical labels and the scarcity of transition state configurations in candidate pools. Active learnin…

  2. arXiv cs.LG TIER_1 English(EN) · Zachary Weller-Davies ·

    Pretrained Model Representations as Acquisition Signals for Active Learning of MLIPs

    Training machine learning interatomic potentials (MLIPs) for reactive chemistry is often bottlenecked by the high cost of quantum chemical labels and the scarcity of transition state configurations in candidate pools. Active learning (AL) can mitigate these costs, but its effecti…

RELATED ENTITIES

RELATED TOPICS