[Nauty-list] Algorithm and special cases

[Hans Georg Schaathun]

Hi all,

I am not sure if this is an appropriate forum to discuss research
ideas, but I take my chances ... if anyone is interested in 
discussing this further, we can possibly do so off-list.

Looking into the algorithm of traces, it seems to me that it can
be sped up for the special case of linear codes.  Linear codes
are represented by bipartite graphs where one vertex set, call it V1,
is much smaller than the other, V2; and as the problem size increases
the V2 will be increasingly overwhelmingly larger.  The interesting thing
is that we are only really interested in a canonical labelling of the
smaller set V1.  The labelling of V2 is discarded when the canonical 
representation is mapped back into a linear code.

It seems to me that it should be desirable, in the linear code case,
to force the algorithm to individualise V1 first, and then possible
to cut the search short once all V1 vertices have been individualised.
Any thoughts on this?

Any comments would be appreciated, and I shall be happy to elaborate
the problem to anyone interested in further discussion.

Thank you very much for reading thus far,
-- 
:-- Hans Georg