Directly generating a user specified K clusters with message passing (Affinity Propagation) 1.0

The Affinity Propagation (AP) clustering algorithmproposed by Frey and Dueck (2007) provides an understandable,nearly optimal summary of a data set. However,it suffers two major shortcomings: i) the number of clusters isvague with the user-defined parameter called self-confidence,and ii) the quadratic computational complexity. When aimingat a given number of clusters due to prior knowledge, APhas to be launched many times until an appropriate settingof self-confidence is found. The re-launched AP increases thecomputational cost by one order of magnitude. We propose an algorithm, called K-AP, to exploit the immediateresults of K clusters by introducing a constraint in theprocess of message passing. Through theoretical analysis andexperimental validation, K-AP was shown to be able to directlygenerate K clusters as user defined, with a negligible increaseof computational cost compared to AP. In the meanwhile, KAPpreserves the clustering quality as AP in terms of thedistortion. K-AP is more effective than k-medoids w.r.t. thedistortion minimization and higher clustering purity (algorithm is described in ICDM'2010).http://www.lri.fr/~xlzhang .

Requirements:	No special requirements
Platforms:	Matlab
Keyword:	Analysis,  Constraint,  Experimental,  Message,  Passing,  Process,  Shown,  Theoretical,  Validation
Users rating:	0/10