Benchmarking Quantum Annealing Controls with Portfolio Optimization

TL;DR

This paper benchmarks quantum annealing controls using portfolio optimization, comparing empirical results from a D-Wave quantum annealer to ground truth, to understand control effects on performance.

Contribution

It provides a systematic evaluation of quantum annealing controls and their impact on optimization accuracy using portfolio problems as a case study.

Findings

Optimal control variations improve success probability.

Reverse annealing shows different error mechanisms.

Certain controls reduce chain breaks significantly.

Abstract

Quantum annealing offers a novel approach to finding the optimal solutions for a variety of computational problems, where the quantum annealing controls influence the observed performance and error mechanisms by tuning the underlying quantum dynamics. However, the influence of the available controls is often poorly understood, and methods for evaluating the effects of these controls are necessary to tune quantum computational performance. Here we use portfolio optimization as a case study by which to benchmark quantum annealing controls and their relative effects on computational accuracy. We compare empirical results from the D-Wave 2000Q quantum annealer to the computational ground truth for a variety of portfolio optimization instances. We evaluate both forward and reverse annealing methods and we identify control variations that yield optimal performance in terms of probability of success and probability of chain breaks.