Originally posted by: srinit34

Hi

can we do something like this:

//set time limit to 3 minutes

ilocplex.solve() //for 3 minutes

//change MIP emphasis

//set time limit to 3 minutes

ilocplex.solve() //for 3 more minutes

Also , what is the relation between MIP emphasis and node selection strategy ( best first, depth first, best estimate first).

Seems to me that depending on choice of MIP emphasis, only certain node selection strategies make sense.

Any other parameter settings to watch out for, while setting MIP emphasis?

Originally posted by: EdKlotz

Yes, you can interrupt an optimization, change the MIP emphasis parameter and continue from where it left off.   CPLEX will adjust the internal parameter settings it changes to account for the new MIP emphasis setting.   Regarding node selection, best estimate and alternate best estimate are used with settings 1 and 4, while best node (with different levels of backtracking) is used with settings 2 and 3.  

I don't know if there's anything else to "watch out" for, but let me mention one feature that in some ways is counterintuitive.   When you set MIP emphasis to 1, CPLEX disables its more expensive heuristics.   The idea is that the tradeoff between faster node throughput giving the less expensive heuristics more chances to succeed compared to the slower node throughput but potential additional solutions found by the more expensive heuristics is a favorable one.    One might expect that setting MIP emphasis to 1 would prompt CPLEX to use all heuristics more aggressively, but that is not the case.