I have been using GLPK to experiment with some Active Constraint methods
for MIP solving. Several of these methods rely on the dual cost returned
by GLPK for individual constraints in a particular node. I am finding
that on the nodes of certain problems, GLPK is claiming that every
active constraint has a dual cost of zero.
As such, I was wondering how GLPK is calculating the dual cost, and if
this high rate of zero dual costs sounds reasonable?
The problem that this is easiest to examine this on is the "pk1.mps"
from the MIPLib2003 problem set. The very first node is returning 15
zero-valued dual costs for the first node.
There's also the chance that I've managed to break GLPK while fiddling
with the code, so any help would be appreciated. I believe I am based
off of the GLPK 4.7 code.
Thanks very much,
--
Chris Kafka
[EMAIL PROTECTED]
_______________________________________________
Help-glpk mailing list
[email protected]
http://lists.gnu.org/mailman/listinfo/help-glpk
