Re: [igraph] Cliques Doubt (Subgraph Weighted Edges)

[Surendar Swaminathan]

Hello Graph Experts,

 I am trying to do subgraph with in a graph those that are highly connected. Clique is not producing what I am looking for.

I am worried because I have 500 nodes and reading the output to find the largest highly connected nodes from clique and move in to less highly connected nodes(Triads). I am not interested in Diads.If a node is selected for one of the subgraph it should not be selected with the next subgraph.

I did do some experiment creating small graph and changing edge weights between nodes.

 I created six nodes A,B,C,D,E,F  

Weights are the number of articles the two nodes have worked together.

  example: Node” A” and  Node” B” have worked only in one article. Node "E"  and "F" have worked 5 articles together. Likewise “A”,”B”,”D” have worked together in 1 article. B,C,D have worked 1 article together. A & C have never worked together.

Created csv file with weight as one attribute. Converted it to Graph object and ran Cliques(graph).

Example 2 :

Node1	Node2	weight
a	b	5
b	d	5
c	d	1
a	d	5
b	c	1
a	c	5
d	e	5
e	h	5

 

[[1]]

[1] 0

 

[[2]]

[1] 1

 

[[3]]

[1] 2

 

[[4]]

[1] 3

 

[[5]]

[1] 4

[[6]]

[1] 5

 

[[7]]

[1] 0 1

 

[[8]]

[1] 0 2

 

[[9]]

[1] 0 3

 

[[10]]

[1] 1 2

 

[[11]]

[1] 1 3

 

[[12]]

[1] 2 3

 

[[13]]

[1] 3 4

 

[[14]]

[1] 4 5

 

[[15]]

[1] 0 1 2

 

[[16]]

[1] 0 1 3

 

[[17]]

[1] 0 2 3

 

[[18]]

[1] 1 2 3

 

[[19]]

[1] 0 1 2 3

 The final iteration is not something I am looking for. Node C should not be included in my clique final answer because Node C and Node D have worked only one article even though Node A and Node C have worked 5 times together.There is no closed loop Node C can fit into. Here is my requirement  A,B,D have published 5 articles together and they seem to be working together more than any other triads and  final clique says 0,1,2,3 .So the final answer is ABD and no other subgraph is correct given the weights and the nodes that has been used for one of the subgraph should not be used for the other subgraph.

Only the iteration #16 gives me proper result.

 [[16]]

[1] 0 1 3

This should be final answer. Is there anyway I can do.

 I hope graph experts can understand. Clique is not the desired solution for me and I have no other idea  work around for this. I am sorry about the long email.

 

On Thu, Apr 2, 2009 at 9:46 AM, Surendar Swaminathan <address@hidden> wrote:

O.k so the better way to proceed is to use weighted graph rather than using multiple links.

On Thu, Apr 2, 2009 at 12:24 AM, Gábor Csárdi <address@hidden> wrote:

2009/4/2 Surendar Swaminathan <address@hidden>:

> Hello Gabor,
>
>   Thanks for the reply.I understand what you are saying betweenness and
> Eigen Values changes between multiple link and single link I have tried and
> they are different.The program works perfectly fine if I input the multiple
> nodes the way I feed.

It is one thing that the functions terminate without an error message
and it is another thing whether the results actually make sense.

Now that I think about it, eigenvector centrality should be good, as
well. It does the same as it would on the corresponding weighted
graph.

Still not sure about betweenness, though.

Best,
Gabor

> With respect to the cliques I will work on them.I tried before I will cross
> check once more.
>
> Nathan
>

[...]

--

Gabor Csardi <address@hidden>     UNIL DGM


_______________________________________________
igraph-help mailing list
address@hidden
http://lists.nongnu.org/mailman/listinfo/igraph-help