[Cbc] 2 buggy examples
John Forrest
john.forrest at fastercoin.com
Mon Feb 22 09:48:50 EST 2016
Gleb,
Should be fixed now.
In knapsack cuts there is an array to say whether variables have been
complemented before trying to create cut. This should be zero before a
row is analyzed, but the test to say whether the array should be zeroed
out after the row is analyzed was correct 99.999999% of time, but .....
John Forrest
On 22/02/16 03:22, Gleb Belov wrote:
> Hi,
>
> attaching 2 examples where cbc reports wrong answers in default
> setting. The 1st one says "infeasible" in 4 seconds, the 2nd delivers
> a wrong optimum 26 in 72 seconds (KUbuntu 15.04, 64-bit, i7-4k).
>
> These models are derived from the original CP models using
> simple&dense unary encodings (i.e., a binary variable for each value
> in the domain). With "-preprocess off" or when the encodings are
> sparse, the wrong answers are not produced.
>
> Gleb
>
>
> _______________________________________________
> Cbc mailing list
> Cbc at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/cbc
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/cbc/attachments/20160222/7e1e2b9d/attachment.html>
More information about the Cbc
mailing list