Number of communities while running Modularity

[Ramonero]

I am wondering if it would be possible to run Modularity by choosing the exact number of communities you want to get.

[admin]

No it's not the way it works.

[Ramonero]

Ok, thanks.
So maybe is a good thing to take on count in the future, at least for me it would be very useful. Im tryng to make selections for the n "more important" nodes in a graph and if i could run modularity by choosing n number of communities it would be great. Now i am changing resolution until i get the number of communities i want.
Greetings

[Ramonero]

Trying to explain a little more...
Lets say we have a big number of nodes, and lets say we have doble number of directed edges... and i want to send some information to only 'n' number of nodes, so i have to choose the 'n' nodes that would make information travel to all nodes in the graph with cost as minimum as possible.
I'll be very grateful for any advice to get this!

[pegerp]

Mathematical properties of modularity don't make it possible to do what you are asking.

You might be more interested in Gephi's Eigenvector Centrality (https://en.wikipedia.org/wiki/Eigenvector_centrality). Centrality of a vertex measures its relative importance within a graph. This is not exactly what you are asking for but might be still more suitable than modularity. Or maybe together with modularity and centrality you'll get something more useful out of the data.

However what you really are after but what is not exactly in the scope of Gephi are flow networks. There exist algorithms that can give answer to your question. Google for flow networks and related issues. In social sciences there are also algorithms that capture the notion of "familiarity" between multiple nodes in a network - not exactly the shortest path or maximum flow but somewhat similar concept - although I can't find the sources now.

You might be able to find other software that has algorithms more suited to your problem. I can't recommend any software but I have a feeling that there are not many programs that do exactly what you'd want it to do. Usually there are small utility programs that compute one thing and then you'd use another software to continue processing the data (eg. Gephi).

[Ramonero]

Thank you very much for your straight useful information. I will look for them and i have the feeling it will be good!
Anycase, if you remember or suddenly find the way, the algoritm, or the utility for helping in my task please let me know. I would save maybe years of work or maybe i will never be able to do it...
Thanks again

[samarth]

Ramonero,

It sounds like you are trying to figure out how to find a good seed set. This paper may be relevant:

A Scalable Heuristic for Viral Marketing Under the Tipping Model
Paulo Shakarian, Sean Eyre, Damon Paulo
http://arxiv.org/abs/1309.2963

[Ramonero]

Does eigenvector centrality take on count if the edges are directed or not?, it seems not and then it is not very useful for me.
About the notion of "familiarity" it sounds interesting but i couldn't find anything related to on the web.
Flow networks or seed sets are not very close to what i am looking for neither. It seems i didn´t explain very good.
What i am working in is closer to concept of small world networks than huge networks.
And also what i am trying to figure out is related to build social networks in a concious way more than studing existing ones.

[pegerp]

About the notion of "familiarity" it sounds interesting but i couldn't find anything related to on the web.
Flow networks or seed sets are not very close to what i am looking for neither. It seems i didn´t explain very good.
What i am working in is closer to concept of small world networks than huge networks.

The algorithm that I couldn't remember earlier is about finding center-piece subgraphs. There is a paper about it at http://www.cs.cmu.edu/~christos/PUBLICA ... 06CePS.pdf. Also the author has given link to Matlab code for the implementation at his website: http://www.cs.cmu.edu/~htong/sel.htm.

Given Q nodes in a social network (say, authorship network), how can we find the node/author that is the centerpiece, and has direct or indirect connections to all, or most of them?

Apparently it's still not exactly what you are looking for. Good luck regardless!

[Ramonero]

The algorithm that I couldn't remember earlier is about finding center-piece subgraphs. There is a paper about it at http://www.cs.cmu.edu/~christos/PUBLICA ... 06CePS.pdf. Also the author has given link to Matlab code for the implementation at his website: http://www.cs.cmu.edu/~htong/sel.htm.

Given Q nodes in a social network (say, authorship network), how can we find the node/author that is the centerpiece, and has direct or indirect connections to all, or most of them?

Apparently it's still not exactly what you are looking for. Good luck regardless!

Yes, it is not exactly what i need but is more closely related. A very interesting paper.
Thanks again!