[Coin-lpsolver] Re: CLP failed on some case

[John J Forrest]

Xiaoping,

Due to limited bandwidth I could only download bad2 and good2..

The coefficients are NOT just bigger.  For example in good2.mps C0014177
has an objective of 296808550 and a -1 in row R0004172 which is an L row.
bad2.mps is the same except the coefficient is -2.14748e9.  With a negative
objective coefficient the problem is unbounded.

An example such as dualcuts.cpp will not have any checking to detect such
problems but as far as I could see the stand-alone clp had no problems
determining that problem was unbounded.

John Forrest


                                                                           
             Xiaoping                                                      
             Tang/Watson/IBM at I                                             
             BMUS                                                       To 
             Sent by:                  coin-lpsolver at list.coin-or.org      
             coin-lpsolver-bou                                          cc 
             nces at list.coin-or                                             
             .org                                                  Subject 
                                       [Coin-lpsolver] Re: CLP failed on   
                                       some case                           
             17.08.2007 15:37                                              
                                                                           
                                                                           
                                                                           
                                                                           
                                                                           





Hi,

I have a question about CLP.

CLP failed on some case that we know there exists a feasible solution.
Please access the testcases in the link (you can download there):

http://yktgsa.ibm.com/~xtang/public/mpstest/

in the directory, good2.mps and bad2.mps are a pair, and good and bad are
another pair.
For good2.mps and bad2.mps, the constraint matrix are the same,
the only difference is, the coefficients of the objective funciton in
bad2.mps
is 10 times as large as the ones in good2.mps.

However, clp successfully finds the feasible solution for good2.mps, but
fails for bad2.mps.
Seems inside clp, some numbers are over-flow.

I tried clp (primal method), dualCuts, and different versions including
Clp-1.4.0, Clp_2006Oct09,
and Clp_2006Oct12.
All give the same result (finish for good.mps and good2.mps, but fail for
bad.mps and bad2.mps).,

Could you please take a look and tell us why?
Any assistance possible, please let me know.?

Thanks a lot,

-Xiaoping

_______________________________________________
Coin-lpsolver mailing list
Coin-lpsolver at list.coin-or.org
http://list.coin-or.org/mailman/listinfo/coin-lpsolver