Auto-regressive Neural Quantum State Sampling for Selected Configuration Interaction

TL;DR

This paper introduces an auto-regressive neural network-guided selected configuration interaction algorithm that efficiently constructs variational subspaces, accelerating ground-state energy calculations in quantum chemistry.

Contribution

It presents a novel SCI algorithm leveraging ARNNs to guide subspace expansion, improving convergence and scalability in quantum chemistry computations.

Findings

Demonstrates improved convergence speed on molecular benchmarks.

Combines neural network representations with classical subspace methods.

Provides a scalable framework for classical and hybrid quantum algorithms.

Abstract

Accurate ground-state energy calculations remain a central challenge in quantum chemistry due to the exponential scaling of the many-body Hilbert space. Variational Monte Carlo and variational quantum eigensolvers offer promising ansatz optimization approaches but face limitations in convergence as well as hardware constraints. We introduce a particular Selected Configuration Interaction (SCI) algorithm that uses auto-regressive neural networks (ARNNs) to guide subspace expansion for ground-state search. Leveraging the unique properties of ARNNs, our algorithm efficiently constructs compact variational subspaces from learned ground-state statistics, which in turn accelerates convergence to the ground-state energy. Benchmarks on molecular systems demonstrate that ARNN-guided subspace expansion combines the strengths of neural-network representations and classical subspace methods, providing a scalable framework for classical and hybrid quantum-classical algorithms.