Algorithms such as Affinity propagation make use of similarity matrix for performing Clustering, Also sparse matrices are preferred esp for large datasets. How does this differs from problem of Community Detection in networks. Since the Similarity Matrix can be considered as equivalent of Adjacency Matrix. (In networks where all nodes in network are of same type.)

asked Jul 01 '10 at 00:20

anon's gravatar image

anon
1473610

reverted Jul 01 '10 at 18:31

Joseph%20Turian's gravatar image

Joseph Turian ♦♦
579051125146

reverted since the answer is good.

(Jul 01 '10 at 18:31) Joseph Turian ♦♦

In addition to what Alexandre Passos said below, remember that the similarity matrix is often just the binary adjacency matrix. Often an explicit point is made of avoiding using any other information. This makes random-walk based methods a huge focus of community detection, whereas I don't think they're all that big in clustering.

(Aug 02 '10 at 12:47) Jacob Jensen

One Answer:

Community detection is just a clustering problem with a similarity matrix, yes. The literature in the area has no problem stating this. It's just that it is a clustering problem with very specific difficulties (and also of interest to a slightly different community), so it has spawned a lot of specific research. Still, most of the methods they use are similar to traditional clustering methods.

This answer is marked "community wiki".

answered Jul 01 '10 at 07:55

Alexandre%20Passos's gravatar image

Alexandre Passos ♦
2554154278421

Your answer
toggle preview

powered by OSQA

User submitted content is under Creative Commons: Attribution - Share Alike; Other things copyright (C) 2010, MetaOptimize LLC.