Originally posted by: SystemAdmin

[hhhxxx said:]

Hi, I have an binary integer linear program that calls solve() multiple times, where each time, a constraint is added to the model to invalidate the previous solution, so that I get the next best solution.  Is there an efficient way of doing this without resolving the problem from scratch each time?  Thanks.
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

[gerardo.trevino said:]

You could save your best solution once you finish each run, then, you could use the SetSolution command available through the Heuristic Callback using your saved solution. Take care of calling your callback just once (at the begining, do nothing otherwise) per solving process. 
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

[prubin said:]

Assuming you have CPLEX 11, you can use the solution pool feature.  By tweaking parameters, you can get it to accumulate several/many/all optimal solutions, several/many/all near-optimal solutions, a population of near-optimal solutions with high diversity, or with the most aggressive settings many/most/all feasible solutions.  (By "solution" I mean integer-feasible solution in all cases.)  The usual caveats apply:  finding all optimal solutions is in general NP-hard (so you can guess how annoying finding all feasible solutions would be); and there can be lots of optimal or near-optimal solutions (so if you're not careful you'll run out of memory).

/Paul

#CPLEXOptimizers
#DecisionOptimization