SCALE-BASED CLUSTERING USING THE RADIAL BASIS FUNCTION NETWORK

This paper shows how scale-based clustering can be done using the radi al basis function (RBF) network (RBFN), with the RBF width as the scal e parameter and a dummy target as the desired output, The technique su ggests the ''right'' scale at which the given data set should be clust ered, thereby providing a solution to the problem of determining the n umber of RBF units and the widths required to get a good network solut ion, The network compares favorably with other standard techniques on benchmark clustering examples, Properties that are required of non-Gau ssian basis functions, if they are to serve in alternative clustering networks, are identified, This work, on the whole, points out an impor tant role played by the width parameter in RBFN, when observed over se veral scales, and provides a fundamental link to the scale space theor y developed in computational vision.