Hi,

I have coded an initial version of k-connectivity,which works but can be
improved.

The link is here https://github.com/Rajesh-Veeranki/sage
-patches/blob/master/k-connectivity.py.

Probably we should be using previously computed results to speeden it.

Thanks
Rajesh V


On Wed, Mar 12, 2014 at 5:48 AM, Rajesh Veeranki <[email protected]>wrote:

> Hi,
>
> I was busy with my academics last week,so i wasn't very active.
> Now, i have written a draft of my 
> proposal<https://docs.google.com/document/d/1SvNraF0x9rNl4Xbpne7l7I_82H_Mz_Hkqt0Ut_Sc4Lc/pub>
> .
> I'm not quite sure as to what is expected ,so please have a look at it and
> suggest what should be added.
> I'll be adding the details of my schedule sooner.
>
> I have a few doubts:
>
> 1.For the minor testing, how do i efficiently find reduced matrix
> representation for each basis.Do i need to use gauss-jordan computation for
> reduced row echelon form or is there some other method?
>
> 2.As per my understanding, the invariants like the tutte polynomial at
> (i,i), canonical partition, support graph can only be used for isomorphism
> testing and not directly for minor testing.Is it so or some of these are
> invariants under deletions and contractions?
>
>
> 3.For the automorphism group computation, when we find that the generated
> group is not the Aut(M), how do we decide what more nonbases to add to the
> graph?
>
>
> Thanks,
> Rajesh
>
>
>
> On Sun, Mar 9, 2014 at 8:02 AM, Rajesh Veeranki <[email protected]>wrote:
>
>>
>> Dear Rudi,
>>
>> On Sun, Mar 9, 2014 at 3:30 AM, Pendavingh, R.A. 
>> <[email protected]>wrote:
>>
>>> Dear Rajesh,
>>>
>>> I'm not sure I understand what your question is, so here's several
>>> answers.
>>> 1) if M is a matroid on E and N its set of nonbases, then the
>>> automorphism group of the hypergraph H=(E, N) equals the automorphism group
>>> of M: a permutation s of E is an automorphism of M if and only if [X is
>>> dependent<=> s[X] is dependent] for each subset X of E with r(M) elements.
>>>
>>
>>
>>> 2) to test if a subgroup G of the permutation group of E contains only
>>> automorphisms of a matroid M on E, it suffices to find a set of generators
>>> of G, and tot test, for each generator g, if g is an automorphism of M.
>>>
>>
>> Ah, i infact forgot about the generators.I was thinking of verifying each
>> element of the subgroup.Thanks, i got the answer.
>>
>> Cheers,
>>> Rudi
>>>
>>> On 8 mrt. 2014, at 16:07, "Rajesh Veeranki" <[email protected]
>>> <mailto:[email protected]>> wrote:
>>>
>>> Dear Rudi,
>>>
>>> Generating nonbases hypergraph that contains the Aut(M) is clear.But
>>> what i'm not clear is how do i verify if the Automorphism group of the
>>> above constructed graph is generated by the automorphism group of the
>>> matroid?
>>>
>>> Can you provide some thoughts on this.
>>>
>>> Thanks
>>>
>>>
>>> On Wed, Mar 5, 2014 at 4:15 AM, Pendavingh, R.A. <[email protected]
>>> <mailto:[email protected]>> wrote:
>>> Dear Rajesh,
>>>
>>> Please call me Rudi.
>>>
>>> As far as i can see,the magic numbers are calculated for each reduced
>>> matrix, displaying each basis.
>>> OK, thanks.
>>>
>>>
>>> Regarding, the automorphism group of a matroid, we should certainly
>>> avoid the full non-bases calculation.Can you please suggest some resource,(
>>> if you have something particular in mind ), for studying the hypergraphs
>>> that are invariants of a matroid?
>>>
>>> I did not have a particular reference in mind. What I was thinking of as
>>> a first attempt was, say, the hypergraph whose edges are all the flats of
>>> rank k, taking O(n^k) time. That particular hypergraph I expect to work
>>> well for linear matroids over small fields, also assuming that the rank of
>>> these matroid is not too big compared to k. But which hypergraph to try
>>> will always depend on the representation of the matroid.
>>>
>>> I am not sure about the BasisMatroid. Here the construction of the
>>> nonbasis hypergraph is relatively inexpensive compared to the size of the
>>> representation, and it seems difficult to come up with any hypergraph that
>>> is faster to compute. But it may still give a speedup to avoid the
>>> construction. I actually have a routine implemented that inspects the
>>> nonbasis hypergraph without constructing it (running through the r-sets),
>>> and which calculates, for each element e, the number of nonbases containing
>>> e. Coloring the groundset elements with this information may cut down the
>>> isomorphism group significantly, since all automorphism must preserve
>>> color. You get this partition by calling M._basis_partition() on a
>>> BasisMatroid. If you get a partition into singletons, the automorphism
>>> group is clearly trivial, and perhaps other outcomes may also be exploited
>>> for a speedup. Also, check out M._basis_partition2(). This does create the
>>> nonbasis hypergraph as a SetSystem, but then finds an equitable partition
>>> for it. In the course of that computation, a trimmed version of the
>>> nonbasis hypergraph is created. This seems a better object to feed to the
>>> graph automorphism group computation, since it's much smaller.
>>>
>>> I'm also reading your paper<http://arxiv.org/pdf/1203.0910v2.pdf> for
>>> understanding the projective invariants of matrices.
>>>
>>> Let me know when something needs clarification. It is a technical story.
>>>
>>> Best,
>>> Rudi
>>>
>>>
>>>
>>> --
>>>
>>>
>>>         Rajesh Veeranki | Senior Undergraduate | C.S.E | IIT Bombay
>>>
>>>
>>
>>
>> --
>>
>>
>> *Rajesh Veeranki** | Senior Undergraduate | **C.S.E** | **IIT Bombay*
>>
>>
>
>
> --
>
>
> *Rajesh Veeranki** | Senior Undergraduate | **C.S.E** | **IIT Bombay*
>
>


-- 


*Rajesh Veeranki** | Senior Undergraduate | **C.S.E** | **IIT Bombay*

-- 
You received this message because you are subscribed to the Google Groups 
"sage-gsoc" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To post to this group, send email to [email protected].
Visit this group at http://groups.google.com/group/sage-gsoc.
For more options, visit https://groups.google.com/d/optout.

Reply via email to