Modified clustering algorithm for projective ART neural network
Roman Krakovsky, Radoslav Forgáč, I. Mokris · 2014
This paper is focused on the description of modified clustering algorithm for PART neural network with multidimensional real-world data. The advantages of the modified algorithm are the elimination of the unassigned patterns into outlier cluster; the ability of algorithm to create projective clusters without generating PART recursive tree; the introduction of centroids and Euclidean metric in the proposed algorithm and finally the small number of learning iterations.