Derivative Informed Learning of Exchange-Correlation Functionals

Arxiv CS.LG (machine learning)

Paper proposes a machine-learned approach to exchange-correlation functionals that uses derivative information to improve accuracy. The method aims to consistently outperform traditional O(N^4)-scaling density functional approximations.

Arxiv CS.LG (machine learning)·Eike S. EberhardLuca A. ThiedeAbdul AldossaryAndreas BurgerNicholas GaoVignesh BhethanabotlaAl\'an Aspuru-GuzikStephan G\"unnemann

7 days ago