Towards Data-Driven Nonlocal Density Functionals: Deep Learning DFT with Attention to approach Chemical Accuracy

Kirill Kulaev, Bogdan Protsenko, Alexander Ryabov, Alexander Guda, Evgeny Burnaev, Vladimir Vanovskiy

Submission Type: I want my submission to be considered for both oral and poster presentation.

Keywords: DFT, density functional theory, neural networks, machine learning, quantum chemistry, exchange-correlation

TL;DR: The model employs an encoder–decoder design with a nonlocal block that refines B3LYP-like exchange–correlation terms (including D3BJ dispersion) through learned smooth enhancement factors while preserving the analytic hybrid structure.

Confirmation Of Submission Requirements: I submit an abstract. It uses the template provided on the submission page and is no longer than 2 pages.

Submission Number: 398