Closed alanctprado closed 4 months ago
1) Gurobi
- Gurobi
sh run_case.sh medium-set 15 --ipsolver="lpsolve" 7.91s user 0.04s system 99% cpu 7.966 total sh run_case.sh medium-set 15 --ipsolver="gurobi" 0.67s user 0.08s system 111% cpu 0.665 total
- Gurobi
sh run_case.sh medium-set 30 --ipsolver="lpsolve" 398.04s user 0.33s system 99% cpu 6:40.16 total sh run_case.sh medium-set 30 --ipsolver="gurobi" 4.39s user 0.28s system 100% cpu 4.663 total
- Gurobi
I think it is pretty clear from a few tests that Gurobi scales drastically better than LPSolve. Nevertheless, it is quite bad for instances with 500 vertices and over (which basically all instances from the exact-set have). The bottleneck is definitely the $N^3$ transitivity constraints. But i think that these tests pave the way forward:
I'll work on testing OR-Tools @alanctprado.
Closed by #26
So far, we have only experimented with LP Solve. We should come up with other IP solvers in order to check the performance differences.
Possible solvers should be listed as comments below.