Originally posted by: RobertFlores

I am accessing Cplex through matlab. For some of my MIP simulations with a node or time limit, Cplex does not achieve a small enough gap. Is there a way to feed the result and node information from the prior optimization back into the solver so that the solver can pick up where it left off, as opposed to redoing the entire optimization again? I have tried to set the x0 initial point as the solution from the prior optimization, but this appears to have no effect on solver performance.

Thank you!

Robert

If CPLEX stops due to a time or node limit then just calling Cplex.solve() again should start from where the previous solve stopped, unless you modified the problem.

If you modify the problem CPLEX will have to drop the search tree since it may have become invalid, so there is no way to reuse that. However, CPLEX retains feasible solutions from the previous run as MIP starts. And if they are still feasible in the second run they will be used as primal solutions.

You can inspect or modify those MIP starts via the Cplex.MipStart property.