Improving the performance of fermionic neural networks with the Slater exponential Ansatz

TL;DR

This paper enhances fermionic neural networks by integrating the Slater exponential Ansatz, leading to faster convergence and more accurate ground-state energy calculations for molecules, with improved efficiency and extrapolation techniques.

Contribution

Introduces a Slater exponential Ansatz for fermionic neural networks, improving convergence and accuracy in quantum chemistry calculations.

Findings

Faster convergence of ground-state energies using the new Ansatz.

Accurate energy estimates with smaller batch sizes via bagging.

Good agreement with CCSD(T) results in the CBS limit.

Abstract

In this work, we propose a technique for the use of fermionic neural networks (FermiNets) with the Slater exponential Ansatz for electron-nuclear and electron-electron distances, which provides faster convergence of target ground-state energies due to a better description of the interparticle interaction in the vicinities of the coalescence points. Analysis of learning curves indicates on the possibility to obtain accurate energies with smaller batch sizes using arguments of the bagging approach. In order to obtain even more accurate results for the ground-state energies, we suggest an extrapolation scheme, which estimates Monte Carlo integrals in the limit of an infinite number of points. Numerical tests for a set of molecules demonstrate a good agreement with the results of original FermiNets (achieved with larger batch sizes than required by our approach) and with results of coupled-cluster singles and doubles with perturbative triples (CCSD(T)) method, calculated in the complete basis set (CBS) limit.