Module AffinityPropagation

Clustering by Affinity Propagation

This clustering algorithm identifies clusters in a set of data items based on a list of similarities between the items. The result is a list of clusters, each cluster being defined by one 'exemplar' (the item that is most representative of the cluster) and by other items. The number of clusters is not specified in advance. Instead, a parameter called 'preference' indicates how likely each item is to be an exemplar. Often it is set to the same value for all items. Low preference values yield few big clusters, whereas high preference values yield many small clusters.

The algorithm is described in: B.J. Frey & D. Dueck, Science 315, 972-976 (2007)