[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Help-glpk] [Fwd: Get the MIp relative gap at the end of the optimizatio
From: |
Andrew Makhorin |
Subject: |
[Help-glpk] [Fwd: Get the MIp relative gap at the end of the optimization] |
Date: |
Tue, 19 Jul 2011 23:29:33 +0400 |
-------- Forwarded Message --------
From: oguyon <address@hidden>
To: address@hidden
Subject: [Help-glpk] Get the MIp relative gap at the end of the
optimization
Date: Tue, 19 Jul 2011 05:45:35 -0700 (PDT)
Hi,
I want to solve a MIP model with two successive stop criterion.
1- I want to find the optimal solution within a specific duration (X ms).
2- If, after X ms, the optimal solution has not been found. I'm looking for
a "good" solution without any time limit.
To do that, I first fix the time limit parameter to X ms (iocp.setTm_lim(X))
and I solve the model. Then, if my solution is not "good" (i.e. the MIP
relative gap is too big): I fix the time limit parameter to a big value, I
fix the MIP gap parameter to a value Y (iocp.setMip_gap(Y)) and I solve the
MIP model a second time.
My problem is that I don't succeed in finding the MIP relative gap at the
end of the first "optimization". The only interesting function that I've
found is : glp_ios_mip_gap(glp_tree arg) ; but I don't have a clue about
what is the tree I have to put in argument.
I'm using glpk 4.45 through a Java environnement (JRE 1.6.0_24).
Here is my full code :
protected double solveApprox(){
int X=5000; Y=0.01, ret;
boolean succes;
glp_iocp iocp = new glp_iocp();
GLPK.glp_init_iocp(iocp);
iocp.setTm_lim(X);
ret = GLPK.glp_intopt(model, iocp);
if(ret!=0)
success = false;
else{
epGap = ????? GLPK.glp_ios_mip_gap(arg0) ????
}
success = (epGap<=relativeMIPGap);
if(success){
resu = GLPK.glp_mip_obj_val(model);
} else {
iocp.setTm_lim(Integer.MAX_VALUE);
iocp.setMip_gap(Y);
ret = GLPK.glp_intopt(model, iocp);
success = (ret == 0);
if(success){
resu = GLPK.glp_mip_obj_val(model);
}
}
return resu;
}
Thank you for your help,
Olivier Guyon
- [Help-glpk] [Fwd: Get the MIp relative gap at the end of the optimization],
Andrew Makhorin <=