Crystalformer: Infinitely Connected Attention for Periodic Structure Encoding

arXiv:2403.11686v1 Announce Type: new
Abstract: Predicting physical properties of materials from their crystal structures is a fundamental problem in materials science. In peripheral areas such as the prediction of molecular properties, fully connected attention networks have been shown to be successful. However, unlike these finite atom arrangements, crystal structures are infinitely repeating, periodic arrangements of atoms, whose fully connected attention results in infinitely connected attention. In this work, we show that this infinitely connected attention can lead to a computationally …

abstract arxiv atom attention cond-mat.mtrl-sci cs.lg encoding however materials materials science networks physics.comp-ph prediction science type