[Dip-tickets] [Dip] #72: Root bound anomaly
Dip
coin-trac at coin-or.org
Wed Mar 16 09:30:31 EDT 2011
#72: Root bound anomaly
-------------------+--------------------------------------------------------
Reporter: vejlin | Type: defect
Status: new | Priority: major
Version: trunk | Keywords:
-------------------+--------------------------------------------------------
Using the MILPBlock example (with CPLEX 12.1 to solve pricing problems) I
sometimes get suboptimal root bounds, that is bounds that are worse than
the simple LP bound.
Steps to reproduce:
1) use MILPBlock example
2) use parameter nodeLimit = 0 (to stop after computing root bound)
3) use I_16.mps instance file (attached to this ticket)
4) use I_16.block block file (attached to this ticket)
5) use BlockFileFormat = List
On my machine this produces the root bound -183.37 which is worse than the
simple LP bound of -182.71 (confirmed with CPLEX and with MILBlock using
an empty block file).
The bounds seem to always be legal but just suboptimal, so from what I can
tell this should affect performance but not the correctness of DIP.
--
Ticket URL: <https://projects.coin-or.org/Dip/ticket/72>
Dip <https://projects.coin-or.org/Dip>
An extensible software framework for implementing decompositon-based bounding algorithms for use in solving large-scale discrete optimization problems.
More information about the Dip-tickets
mailing list