[Dip] cut violation, strange termination etc

[Shahin Gelareh]

Dear all

I have (from among many sets of constraints) a set of constraints in the
pricing problem with a large cardinality.
If I solve the problem for the small size instances, it works fine.

As soon as it gets larger it runs into difficulties.

I thought moving those set of constraints to the function "generateCut" and
add them only upon violation.

Special care has been taken to avoid duplicating the already added cut.
The issues are the followings:

1)  for my minimization problem, gUB drops below the optimal
2) at some point DIP says that a solution resulted by the convex
combination is actually a solution that we have fixed to zero before and
this is the reason for crash.
3)  code crashes as follows i guess due to some violation.

 + -1 x[5210] + -1 x[5299] + -1 x[5388] + -1 x[5477] + -1 x[5566] + -1
x[5655] + -1 x[5745] + 1 x[607] lb: -INF ub: 0 vio: 0
decompCut->getViolation(): 0

 + -1 x[809] + 1 x[2338] lb: -INF ub: 0 vio: 0
decompCut->getViolation(): 0

 + -1 x[2246] + -1 x[2338] + -1 x[2427] + -1 x[2517] + -1 x[2699] + 1
x[3953] lb: -INF ub: 0 vio: 0
decompCut->getViolation(): 0

 + -1 x[3681] + -1 x[3775] + -1 x[3864] + -1 x[3953] + -1 x[4042] + -1
x[4132] + -1 x[4314] + 1 x[5566] lb: -INF ub: 0 vio: 0
decompCut->getViolation(): 0
Calling cut generator: cliques
Calling cut generator: cliques
Calling cut generator: flow-covers
Calling cut generator: knap-covers
Calling cut generator: mixint-round
Num clique     cuts= 5
Num odd-hole   cuts= 3
Num flow-cover cuts= 0
Num knap-cover cuts= 0
Num mixed-int  cuts= 0
Num gomory     cuts= 0
D-ALGO   : 14.3     [CPU: 14.33   ]  --- generateCuts() ----------> funcT =
0.181
D-ALGOPC : 14.3     [CPU: 14.3    ] <--- addCutsToPool() ---------
CUT 1 is not violated! Not adding to pool.
CUT           Hash = 0_L_2904_-1_6140_1_

 + -1 x[2904] + 1 x[6140] lb: -INF ub: 0 vio: 0
Assertion failed: 0, file ..\..\..\..\src\DecompAlgoPC.cpp, line 1234
Press any key to continue . . .



I appreciate the comments.

regards
Shahin
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/dip/attachments/20140317/9d25adc4/attachment.html>