Neural-network-supported basis optimizer for configuration interaction problems
We are pleased to announce the publication of our new paper in Physical Review B. Our work introduces a neural-network approach to optimize the selection of Slater determinants in configuration interaction (CI) calculations for strongly correlated quantum systems. Using a neural classifier and active learning, the method efficiently identifies the most relevant determinants for the ground-state wave function, improving computational compactness without compromising accuracy. The approach is benchmarked against established truncation schemes and demonstrates substantial gains in efficiency and scalability for quantum cluster models.
Authors: Pavlo Bilous, Louis Thirion, Henri Menke, Maurits W. Haverkort, Adriana Pálffy, and Philipp Hansmann.
Published: Phys. Rev. B 111, 035124 (2025)
DOI: https://doi.org/10.1103/PhysRevB.111.035124