[Coin-discuss] Total unimodularity

[Spencer Fung]

Hi all,

I am working on a 0-1 Integer programming problem, which in the form of Ax =
b. The problems I am dealing with has a special structure in the matrix A.
It definitely not a totally unimodular (TU) matrix, since we cannot find a
integer solution by Simplex. But when I do branch-and-bound (branching on
fractional value), after few iterations, an integer solution can be found by
Simplex method. I suspected that the matrix A is not TU, but it is very
close to TU. So, I have two questions,

1. Given any {0,1}-matrix A, can we measure how close of A to be a TU
matrix?
2. Any fast implementation to check the total unimodularity?

Thank you.

Cheers,
Spencer

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/coin-discuss/attachments/20051202/0cc992f2/attachment.html>