Statistical quality assessment of Ising-based annealer outputs

TL;DR

This paper presents a statistical testing method to evaluate the quality of outputs from Ising-based annealers, using sample data to estimate ground-state probabilities and validate device performance.

Contribution

It introduces a novel statistical approach for assessing quantum annealer outputs based solely on sample data, enabling primary validation of their performance.

Findings

Method effectively estimates ground-state energy and temperature.

Bootstrap resampling provides reliable parameter estimates.

Demonstrated on real quantum annealer samples.

Abstract

The ability to evaluate the outcomes of quantum annealers is essential for such devices to be used in complex computational tasks. We introduce a statistical test of the quality of Ising-based annealers' output based on the data only, assessing the ground state's probability of being sampled. A higher probability value implies that at least the lower part of the spectrum is a part of the sample. Assuming a plausible model of the univariate energy distribution of the sample, we express the ground-state energy and temperature as a function of cumulants up to the third order. Using the annealer samples, we evaluate this multiple times using Bootstrap resampling, resulting in an estimated histogram of ground-state energies and deduce the desired parameter on this basis. The approach provides an easily implementable method for the primary validation of Ising-based annealers' output. We demonstrate its behavior through experiments made with actual samples originating from quantum annealer devices.