Hacker News Reader

I have a dumb question: how long will it take before this result becoming a pratical MIP solver beating SCIP or gurobi?

Don't forget about the HiGHS solver [1]. MIT licensed and getting to the point where it's outperforming SCIP on the Mittelmann benchmarks [2].

[1]: https://github.com/ERGO-Code/HiGHS

[2]: https://mattmilten.github.io/mittelmann-plots/

HiGHS is more of an alternative to Bonmin and Minotaur than Couenne and SCIP. In my experience though the presolvers in SCIP are extremely dangerous, and it's easy to end up with a local optimum even when that isn't your goal.

Couldn't either of those implement this algorithm?

Reply on news.ycombinator.com