I’m interested in understanding some of the details in your heuristic. Can you forward me some references or a short description of what the heuristic does? Much appreciated. Thanks!

]]>https://www.ibm.com/developerworks/community/blogs/jfp/entry/d_wave_vs_cplex_comparison_part_2_qubo?lang=en

with results on your three instance sets (439, 502, and noextfield). CPLEX finds the same objective values as your solver on all instances, which is a pretty strong cross validation. Timings are also interesting, no solver dominates the other. ]]>

If so, then perhaps we should adjust the D-Wave time down by some factor, but I’d be surprised if it made more than a factor of 100 difference, in which case the conclusion would surely be similar.

Re QUBO time, I tried my best to use the definitions in your paper. I used the smaller figure for D-Wave running time from section 4, rather than the flat 490ms from section 3.2. I don’t see the two definitions in your paper that make D-Wave come out faster.

]]>