Proprietés asymptotiques en classification (Convergence d’un schéma d’approximation stochastique). (French) Zbl 0566.62049

Lemaire, J.

Data analysis and informatics III, Proc. 3rd Int. Symp., Versailles 1983, 345-358 (1984).

[For the entire collection see Zbl 0559.00011.]
The asymptotic stability of the solutions of clustering problems arises in data analysis when the number n of the observations to be classified is very large. In order to produce effective approximations of the solutions of clustering problems, the convergence of a recursive algorithm has been studied dealing with centroid methods. This convergence has been proved with the help of stochastic approximation results adapted to the clustering framework.

MSC:

62H30	Classification and discrimination; cluster analysis (statistical aspects)
62L20	Stochastic approximation

Keywords:

asymptotic stability; clustering; approximations; convergence of a recursive algorithm; centroid methods

Citations:

Zbl 0559.00011