On 3/7/07, François Duranleau <duranlef@iro.umontreal.ca> wrote:

On Wed, 7 Mar 2007, Stephan Diederich wrote:

> On 3/6/07, François Duranleau <duranlef@iro.umontreal.ca> wrote:
>>
>> It seems there is an inconsistency (to me) between color tagging of
>> vertices in edmund_karp_max_flow and kolmogorov_max_flow algorithms.

[snip]

Hm, I am confused. I compiled and run your example, and yes, the results
are identical... But with the following example, I get inversed colors,
and the flow returned by edmund_karp_max_flow is 0. What am I doing wrong?
I use gcc-4.0.3 and Boost head cvs. I also tried with 1.33.1+kolmogorov,
same thing.

First, thanks for the example. I could reproduce this here. But I don't have the time to look into this deeper before the weekend.
One thing I've seen is that edmunds_karp requires the reverse-edge weights to be zero. But changing it didn't make it better.
Changing the edge_weights from max() to something smaller made edmunds_karp at least return the same flow as kolmogorov_mf does.

I'll look into it at the weekend.

Cheers,
Stephan