A Relative Approach to Comparative Performance Analysis for Quantum Optimization

We discuss a small study on how to compare the performance of various solving techniques for quadratic unconstrained binary optimization (QUBO). Since well-known metrics are seldomly applicable, we suggest comparing the relative performance, i.e., how much the quality of solution (compared to other solutions of the same solver) for a QUBO shifts between different solving techniques. We propose looking for big shifts systematically for an empirical complexity analysis. Code is available at github.com/thomasgabor/gecco- relative.