Sampling asymmetric open quantum systems for artificial neural networks

TL;DR

This paper introduces a hybrid sampling strategy that explicitly accounts for asymmetries in open quantum systems, significantly improving scalability and convergence of neural network models compared to traditional methods that struggle with asymmetries.

Contribution

The authors develop a novel hybrid sampling approach tailored for asymmetric open quantum systems, enhancing neural network applicability beyond symmetric cases.

Findings

Achieves faster convergence times in asymmetric systems

Demonstrates high scalability for complex quantum setups

Outperforms traditional symmetric-based sampling methods

Abstract

While established neural network approaches based on restricted Boltzmann machine architectures and Metropolis sampling methods are well suited for symmetric open quantum systems, they result in poor scalability and systematic errors for setups without symmetries of translational invariance, independent of training parameters such as the sample size. To overcome this representational limit, we present a hybrid sampling strategy which takes asymmetric properties explicitly into account, achieving fast convergence times and high scalability for asymmetric open systems, underlining the universal applicability of artificial neural networks.