The Evolving Tree: a new kind of self-organizing neural network

Abstract

This paper presents a new neural network system called the Evolving Tree. This network resembles the Self-Organizing map, but deviates from it in several aspects, which are desirable in many analysis tasks. First of all the Evolving Tree grows automatically, so the user does not have to decide the network’s size before training. Secondly the network has a hierarchical structure, which makes network training and use computationally very efficient. Test results with both synthetic and actual data show that the Evolving Tree works quite well. 1