01021960

j

01021960 Den{\oe}ux, Thierry Govaert, G\'erard A nonparametric clustering algorithm. (Un algorithme de classification automatique non param\'etrique.) C. R. Acad. Sci., Paris, S\'er. I 324, No.6, 673-678 (1997). 1997 Elsevier, Paris FR clustering algorithm stable partition nonparametric discrimination rule Hopfield connectionist model energy function stationarity classification likelihood maximisation

doi:10.1016/S0764-4442(97)86988-1

Summary: We present a new clustering algorithm based on the search for a stable partition with respect to a nonparametric discrimination rule. A formal link between this method and the Hopfield connectionist model is shown, allowing for the definition of an energy function whose decrease at each iteration guarantees the stationarity of the sequence of partitions obtained. A link between this approach and the principle of classification likelihood maximisation is also studied.