Daniela Scherer dos Santos,  Ana L. C. Bazzan. 

Traditional clustering methods have been usually developed in a centralized fashion; additionally, they need some hints about the target clustering (e.g. number of clusters, expected cluster size, or minimum density of clusters). However this does not meet a typical necessity in multiagent scenarios that is self-organization without central control. In this work we use a clustering algorithm that is inspired by swarm intelligence techniques, is distributed, and does not require any initial hint about the number of clusters. Tests using two applications – one employing typical public datasets and one in a dynamic scenario – show the formation of groups in a distributed way with a performance that is comparable to that achieved using centralized approaches.

http://www.lbd.dcc.ufmg.br/colecoes/enia/2009/038.pdf

Processando consulta.