Originally posted by: Namal

Hello,

I wrote a program that finds me a set of two binary variables for which my problem becomes infeasible if I fix them. It looks like this.

1. set all variables and constraints

2. add the two fixed variables as constraints

3. solve the problem

4. remove the two constrains

5. repeat with different pairs of variables from 2

The problem is that it becomes really slow and I wanted to ask if there is a way to use solve function in such a way that it doesn't have to compute everything all over again?

Originally posted by: BoJensen

What do you actually mean by fixing the two variables with a constraint ? If you wish to fix two binary variables to their either lower or upper bound, then it's better and faster to just change the bounds. Regarding reoptimization, then MIP is a bit different than LP. In LP you most often have efficient restarts, which is not the case in MIP.

Originally posted by: Namal

Hey Bo, it is actually a LP, all the variables can be floats, exept the two that I fix either to zero or one. Sorry for that misleading description. I set the bounds like that

var.add(IloNumVar(lp, 0, 1));

So you say that I should just change the bounds. I have found this function to do so setBounds. If I set lower bound and upper bound the same, does it stay LP or is it already MIP?

Originally posted by: BoJensen

Take a look here for how to modify a problem. If you fix all integer variables in a MIP,  then it becomes a LP,  but fixing variables in a LP can not make it a MIP.