Experimental analysis of quantum annealers and hybrid solvers using benchmark optimization problems

TL;DR

This paper evaluates quantum annealers and hybrid solvers on benchmark problems like HCP and TSP, introducing a new matrix formulation, comparing system efficiencies, and analyzing constraints for problem encoding.

Contribution

It presents a novel matrix formulation for HCP and TSP Hamiltonians, compares different D-Wave systems, and analyzes the constraints needed for problem encoding.

Findings

D-Wave Advantage_system4.1 outperforms Advantage_system1.1 in efficiency and solution quality.

Hybrid solvers reliably produce valid solutions for large problems up to 120 nodes.

Incomplete graphs require more qubits and constraints than complete graphs, affecting problem encoding.

Abstract

This paper studies the Hamiltonian Cycle Problem (HCP) and the Traveling Salesman Problem (TSP) on D-Wave's quantum systems. Initially, motivated by the fact that most libraries present their benchmark instances in terms of adjacency matrices, we develop a novel matrix formulation for the HCP and TSP Hamiltonians, which enables the seamless and automatic integration of benchmark instances in quantum platforms. our extensive experimental tests have led us to some interesting conclusions. D-Wave's {\tt Advantage\_system4.1} is more efficient than {\tt Advantage\_system1.1} both in terms of qubit utilization and quality of solutions. Finally, we experimentally establish that D-Wave's Hybrid solvers always provide a valid solution to a problem, without violating the QUBO constraints, even for arbitrarily big problems, of the order of $120$ nodes. When solving TSP instances, the solutions produced by the quantum annealer are often invalid, in the sense that they violate the topology of the graph. To address this use we advocate the use of \emph{min-max normalization} for the coefficients of the TSP Hamiltonian. Finally, we present a thorough mathematical analysis on the precise number of constraints required to express the HCP and TSP Hamiltonians. This analysis, explains quantitatively why, almost always, running incomplete graph instances requires more qubits than complete instances. It turns out that incomplete graph require more quadratic constraints than complete graphs, a fact that has been corroborated by a series of experiments.