We experimentally study the performance of a programmable quantum annealing processor, the D-Wave One (DW1) with up to 108 qubits, on maximum SAT problem with 2 variables per clause (MAX 2-SAT) problems. We consider ensembles of random problems characterized by a fixed clause density, an external parameter which we tune through its critical value in our experiments. We demonstrate that the DW1 is sensitive to the critical value of the clause density. The DW1 results are verified and compared with akmaxsat, an exact, state-of-the-art algorithm. We study the relative performance of the two solvers and how they correlate in terms of problem hardness. We find that the DW1 performance scales more favorably with problem size and that problem hardness correlation is essentially non-existent. We discuss the relevance and limitations of such a comparison.Keywords: adiabatic quantum computation, quantum annealing, computational complexity
IntroductionAdiabatic quantum computation (AQC) is a model of solving computational problems, in particular hard optimization problems, by evolving a closed system in the ground state manifold Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
P f ad, represents the solution to the computational problem. AQC has been proven to be polynomially equivalent to standard, closed-system, circuit model QC [3][4][5][6][7], but so far it is unclear whether this equivalence extends to the open system, non-zero New J. Phys. 16 (2014) 045006 S Santra et al 2 New J. Phys. 16 (2014) 045006 S Santra et al 3 5 In more detail, the McGeoch and Wang (MW) study [35], working with the DW2, used a 439 qubit subgraph of Chimera and considered three problems: (1) Chimera-structured QUBO instances (this is actually an ensemble of uniform samples from the Ising model on Chimera with ∈ − J h , { 1,1} ij j[36]), (2) Weighted Max 2-SAT, (3) the Quadratic Assignment Problem. Their main conclusion is that in their experiments the DW2 (together with a software layer called Blackbox) outperformed the software against which it was tested. In the case of problem (1), the DW2 is reported as outperforming its nearest rival (CPLEX), amongst those tried, by a factor of 3600. The times recorded by MW were for the first point that CPLEX found the optimal solution, and not the time at which it proved it optimality. However, several researchers have reported classical implementations for all three problems which outperform the DW2 and in particular the MW benchmarks [36,37]. Our ensemble of MAX 2-SAT problems differs from the weighted MAX 2-SAT problems considered by MW, since we used uniform weights with the additional constraint of a fixed clause density. In addition, unlike MWʼs case (2), our MAX 2-SAT problem ensemble inherits the native Chimera graph structure along with the connectivity contraints of the processor by design, which eliminates the ...