NoRel Heuristic

Hey Folks:

I have a MIP that’s super large and hard to solve. My guess is that the polytope is very large and very flat. I’m using Gurobi and I’ve found that the NoRel Heuristic works really well on the problem. Like miles better.

I have also been playing with the Heur option (fraction of time solver spends on heuristics) and see that increasing the fraction leads to better solve times.

Ok, so here’s my question - is NoRelHeur just the same heuristic that would be used during the MIP phase of solving? Would I be better to move some time out of NoRelHeur and into the solver, but bumping up the fraction of time devoted to heuristics?

The NoRel heuristic is different from the other heuristics run by Gurobi. As the name is trying to suggests, this heuristic does not use information from the continuous relaxation of the MIP, but instead takes a completely different approach to looking for feasible solutions.

Thus Gurobi’s norelheurwork and norelheurtime parameters act independently of the heurfrac parameter. You’ll need to experiment to see what the best combination of these settings is for your particular application.