SCIP's default today is still relpscost, but advanced by @timoberthold's use of CP stats (https://t.co/YOZg79V72t), Gregor Hendel's use of significance tests on pseudo cost distributions (https://t.co/nNNqr4lY1x), or @geraldgamrath's use of propagation (ht
584 followers
309 followers
@mluebbecke @BarryOSullivan @ecmiindmath I think an interface layer between ML and optimization is missing. The interface will naturally emphasize the separation of learning and optimization. I have not seen much attempt in this direction. An example: htt
397 followers
RT @localsolver: Nice paper by @posetpartition about Partial Latin Square Extension with benchmarks against #CPLEX and #LocalSolver https:/…
4,033 followers
Nice paper by @posetpartition about Partial Latin Square Extension with benchmarks against #CPLEX and #LocalSolver https://t.co/Vanfa51PUk