Quantum vs Classical Solvers: Benchmarking QUBO Performance

Quadratic Unconstrained Binary Optimization (QUBO) problems describe decisions with binary variables. For a matrix Q, the problem is

minimize xᵀQx over x ∈ {0,1}ⁿ.

A solver searches for the binary vector x with the lowest objective value.

This project compares recorded results from a D-Wave Advantage quantum annealer with classical solvers on the same QUBO instances. The benchmark contains 1,912 instances, solver implementations, and per-instance results.

Project Highlights

  • Benchmark data: Four evaluated instance families cover graph-structured, random, spin-glass, and QPLIB problems.
  • Solver code: The repository contains a self-contained classical reference solver, legacy interfaces for Gurobi and Geno, and a D-Wave interface.
  • Recorded results: Of 1,856 instances with results from both approaches, the classical reference solver returned a lower objective value for 1,280 and the same value for 576.

The benchmark instances, solvers, and results are available in the QUBO benchmark repository.