Efficient implementation of configuration interaction and selected CI Using reinforcement learning

Abstract

In order to account for the missing electron correlation from the Hartree- Fock method, the Configuration Interaction method uses a variational wave function that is a linear combination of configuration state functions (CSFs) built from spin orbitals. Full CI scales unfavorably with the number of or- bitals and electrons relative to other orbital methods. Hence, implementa- tion of CI is done in Rust using numerous Code Optimizations, making this implementation extremely efficient. To deal with an even bigger systems, se- lected Configuration Interaction is explored, and a very promising Reinforce- ment Learning-based method has been implemented and improved further for excited states energies.