New ADAPT model uses Transformer for long-range atomic interactions

Researchers have developed ADAPT, a new machine learning force field that bypasses graph neural networks for modeling atomic interactions. This approach directly uses atomic coordinates and a Transformer encoder to capture all pairwise interactions, aiming to improve the representation of long-range forces. In tests on silicon point defects, ADAPT demonstrated a significant reduction in force and energy prediction errors compared to existing graph-based models, while also being computationally less expensive. AI

IMPACT Offers a more efficient and accurate method for simulating material properties, potentially accelerating materials discovery.

RANK_REASON New research paper detailing a novel machine learning model for materials science. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• Evan Dramko
• Graph Neural Networks (GNNs)
• silicon
• Transformer encoder

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →