Problem with bounds produced by Alps at node 0

[svigerske]

Issue created by migration from Trac.

Original creator: kmartin

Original creation time: 2010-08-23 01:39:39

I notice that for a number of problems that I am running ALPS reports the following lower bound at node 0

Alps0240I Processed 0 nodes, has 0 nodes, best relaxed 1e+75, best feasible 16

This is for a minimization problem where all problems have lower bounds of 0 and upper bounds of 1. How can the best relaxed value exceed the best feasible value for a minimization?

Note also the following odd number being reported for the depth

Alps0260I Quality of the best solution found: 16 ; depth 2147483647

Thanks

[svigerske]

Comment by @tkralphs created at 2010-08-23 01:43:09

This is most likely a problem in ALPS, not DIP, but I will take a look at it.

[svigerske]

Comment by @mgalati13 created at 2010-08-24 20:01:46

These are probably for cases that solve in the root node and the overall optimal is probably being found before Alps is updated. Either way, Alps should init the best relaxed at -inf, not inf. Ted - can you correct that.

I am not sure about the depth problem. My guess is that it is another initialization issue. It seems to be init'd at INT_MAX instead of 0.

[svigerske]

Comment by kmartin created at 2010-09-03 01:35:31

To see this behavior run the osDip example and in the parameter file osdip.parm use the test case

#first simple plant location problem
OSiLFile = osilFiles/spl1.osil
#setup constraints as blocks
OSoLFile = osolFiles/spl1.osol

You will get the

best relaxed 1e+75, best feasible 16

and indeed the lp relaxation is the optimal integer solution