Is my problem a fit?
When classical is still the right answer
An honest checklist before you reach for quantum hardware.
5 minute read
Classical solvers are extremely good, and they keep improving. Part of being problem-first is knowing when not to use new hardware.
Classical is likely better when
- A commercial MILP solver already reaches an acceptable gap in acceptable time.
- The problem is continuous and convex. Gradient methods and interior-point solvers are hard to beat.
- The bottleneck is I/O, data movement or memory rather than search.
- You need certified optimality proofs that only exact classical methods provide today.
Hybrid is worth trying when
- The problem is discrete, highly constrained and the classical solver stalls as size grows.
- A good-enough answer quickly is worth more than a proven optimum slowly.
- You need many diverse near-optimal solutions, not just one.
Whatever you try, compare it against the best classical baseline you can find, not a weak one. The Benchmarking honestly track covers how.
A speedup against a poorly tuned classical baseline is not a speedup. Reviewers and other members will ask.
Key points
- Strong classical solvers are the baseline to beat.
- Hybrid approaches are most promising for large, discrete, constrained problems.
- Always benchmark against a well-tuned classical run.
Check your understanding
Answer the questions above to finish.