Open gameveloster opened 2 years ago
The Obj in the log is from the LP solver, so it's about solving the relaxation. The gap is only checked after the LP relaxation has been solved. In this case, it looks like the solution of the LP relaxation is also feasible for the problem itself, so the gap is zero.
If your problem is actually an LP, then the answer is that the ratioGap doesn't apply here. That's for MIPs or other problems that are solved by methods where primal and dual bounds are computed separately.
The objective increases because Clp has decided to start at an infeasible point with objective value 0 and then works its way towards a feasible point. The more feasible it gets, the worse becomes the objective function value.
Hello, I'm running
cbc
using Python'spyomo
package with theratioGap
option set to a high value of0.9
, so I expect the solver to terminate very quickly after starting.However, it appears to ignore the
ratioGap
option and solve it to optimality. Is this expected, or do I have an incorrect understanding of whatratioGap
is supposed to do?ratioGap
also appears to be ignored when I try other values like0.99
,0.5
,0.1
and0.01
. In all these cases, the number of iterations ran is the same.I'm using CBC 2.10.5, pyomo 6.4.1 and Python 3.10.4, Ubuntu 20.04.
Thank you for any help!
Logs:
Problem File: tmpqsr5ldk_.pyomo.lp
Solution File: tmpqsr5ldk_.pyomo.soln
PS: I'm sort of curious why the obj values in the logs are increasing when this is a minimization problem