Neural-network quantum state for exact diagonalization in CAS-CI calculations
This work published on JCTC utilize Boltzmann machine (BM) architectures as an encoder of ab initio molecular many-electron wave functions represented with the complete active space configuration interaction (CAS-CI) model. This ansatz termed the neural-network quantum state or NQS, first introduced by Carleo and Troyer in their seminal paper, is utilized for finding a variationally optimal form of the ground-state wave function on the basis of the energy minimization. In addition to RBMs that was implemented in the original paper, we further introduced fully connected BMs, and higher-order BMs to explore convergence to global minima afforded by their concave log-liklihood functions.
The algorithm was implemented with an in-house program suite ORZ that performs the quantum chemical calculations. A snippet of the code I wrote for this work is provided in this repository, for showcasing purposes.