Grunnleggende konsepter
Efficient graph transformers enable accurate crystal material property prediction.
Sammendrag
Efficiently predicting crystal material properties is crucial for various applications. Traditional methods are costly and time-consuming, leading to the adoption of machine learning models. However, existing crystal graph representations struggle with capturing complete geometric information, hindering accurate predictions. This paper introduces SE(3) transformers, ComFormer variants, to address these challenges. By utilizing lattice-based representations and invariant/equivariant descriptors, the proposed models achieve state-of-the-art predictive accuracy across different crystal benchmarks.
Statistikk
Experimental results demonstrate the state-of-the-art predictive accuracy of ComFormer variants.
The complexity of the proposed models scales to large-scale crystal datasets with O(nk).
Previous methods fail to capture periodic patterns of crystals and maintain geometric completeness.
Sitater
"Crystal structures are characterized by atomic bases within a primitive unit cell that repeats along a regular lattice throughout 3D space." - Content
"Constructing graphs that effectively capture the complete geometric information of crystals remains an unsolved and challenging problem." - Content