Re: [igraph] Hierarchical Clustering

[Shrijeet Paliwal]

Gabor,

Thanks for taking time. Input, in my opinion is just one network - weighted graph of tags. I am not sure how we can make use of the second network (the hierarchy) and how can we come up with it, it was not in my mind. Actually coming to think of it this is something which is desired as out put (In a way)

>  A soft clustering? How would you define it?

 

See the attached picture. The tag channel could be present in more than one clusters again through a way of hierarchy. Its opposite of flat clustering with which we are familiar with - each tag go to one cluster period.

--
Shrijeet
to iterate is human,to recurse divine.

On Sat, Jan 3, 2009 at 2:17 PM, Gábor Csárdi <address@hidden> wrote:

Shrijeet, there is no such algorithm in igraph, AFAIK.

But I am not completely sure what exactly you want to do. I understand
that the input of your algorithm is 1) a weighted graph of tags and 2)
a hierarchy of tags; basically two separate networks. What would be
the output? A soft clustering? How would you define it?

Best,
Gabor

On Sat, Jan 3, 2009 at 7:48 PM, Shrijeet Paliwal
<address@hidden> wrote:
> Hi Everyone,
>
> I am working on a project in which I am trying to cluster tags (tags
> associated with blog posts, images, bookmarked URLs etc.). I have used few
> implementations in igraph one of them was "igraph_community_fastgreedy"
> which lets me cluster my weighted graph where nodes are tags and edges are
> similarity measure between two tags.
>
> The output we get using these approaches in hard clusters where a tag will
> appear in one and only one cluster. Now I have a situation where I want to
> produce a hierarchical clustering where some dominant tags are at higher
> levels in  hierarchy and hence appaear in one or more clusters where as
> specific and less generic tags are fal leaves which shall appera in just one
> cluster.
>
> Is there any algorithm in igraph library which you think might lead me some
> where? Any other directions or corrections in the way I have framed my
> question will be deeply appreciated.
>
> Thank you guys.
> --
> Shrijeet
>

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



--
Gabor Csardi <address@hidden>     UNIL DGM


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

channel.jpeg
Description: JPEG image