ABSTRACT
This chapter contains some classes of prototype-based partitional clustering algorithms. In order to obtain more flexible algorithms, various adaptive techniques are considered. The obtained adaptive algorithms deal with objective function minimization. The objective functions are defined as the sum of squared distances of points from the cluster prototypes. In the case of point prototypes, the use of different metrics allows the detection of cluster shapes ranging from spherical to elongated clusters. The use of adaptive and local distances considered in Chapter 9 permits us to detect local variations of cluster geometry. In this chapter, we will consider an approach in which the matrix giving the metric is itself variable.
