Closed andreaslundell closed 2 years ago
It definitely fixed very many. I only have three instances return with a wrong dual bound:
----------------------------+------+---------- Original -----------+----------------+----------------+------+---------+--------+-------+-------
Name | Type | Conss | Vars | Discr%|Nonlin%| Dual Bound | Primal Bound | Gap% | Iters | Nodes | Time | Status
----------------------------+------+-------+-------+-------+-------+----------------+----------------+------+---------+--------+-------+-------
ex3pb MINLP 32 33 24.2 5.0 130 130 0.0 70 23 0.4 fail (opt=68.0097) (1/1)
lip MINLP 84 61 85.2 14.1 5435267.88 5435267.88 0.0 1 0 0.0 fail (opt=5.68507e+06) (1/1)
risk2bpb MINLP 581 464 3.0 0.1 -56.8208726 -56.8208726 0.0 6 0 0.1 fail (opt=-55.8761) (1/1)
This should fix many of the issues you found @svigerske (as well as some other ones) for nonconvex problems in GAMS.
I have tested it, and it seems to work on most of the list of problems you sent me, but I would appreciate if you could rerun your tests as well.