Warm Start for LP in gurobi (Python)
Awaiting user inputI have an LP model which I optimize multiple times, each time it is modified by the addition of a constraint that makes the previous optimal solution infeasible. Gurobi automatically warm-starts the LP and by doing so the final optimal solution is not what it should be. As soon as I deactivate the warm start I can successfully find the optimal solution. This should not happen since, as I understand, warm starting the model should make it converge faster and it should not interfere with the optimal solution. What I was wondering is if it could be due to the fact that the basis/solution with which the warm start is performed has become infeasible for the model at the next iteration, or if should look somewhere else for an explanation.
-
Hi Federica,
This in theory should not happen. It could happen in practice if the model is ill-conditioned or causing numerical issues for the solver. Do you see any warnings or signs in the log file that there are numerical issues? What happens if you try different values for Seed, do you still observe that same behavior?
- Riley
0
Please sign in to leave a comment.
Comments
1 comment