Originally posted by: PierreNP

Hi,

I 've implemented in a solve callback that in each node of the mip branch and cut gives a near lp optimal solution that seems to be good enough to evaluate the node bound.

In the solve callback, cplex lp solve call start from this near optimal solution and then takes too much time to reach the lp optimum solution.

Is there a way to tell cplex to be less precise in the lp solve in a mip solve callback?

Best regards,

Pierre. 

Originally posted by: RWunderling

Have you considered relaxing the simplex tolerances?

Roland

Originally posted by: PierreNP

Thanks for the answer!

Yes I did that but without success.

It seems that cplex is searching a basis and that takes much time.

There is a method setBasisStatuses. I Don't know if I can use it in the solve callback to inform cplex about the basis solution statuses, maybe it could help.

edit: It seems setBasisStatuses is not in the solve callback api so I presume it can't be used in the solve callback.

will it help if in the callback, I put a primal and a dual solution in the setStart method?

regards,

Pierre.