Self-organizing neural networks for data projection

Abstract

In this paper we present a nonlinear projection method for visualizing high-dimensional data as a two-dimensional scatter plot. The method is based on a new model of self-organizing neural networks. An algorithm called “double self-organizing feature map” (DSOM) algorithm is developed to train the novel model. By the DSOM algorithm the network will adaptively adjust its architecture during the learning phase so as to make neurons responding to similar stimulus be clustered together. Then the architecture of the network is graphically displayed to show the underlying structure of the data. Two data sets are used to test the effectiveness of the proposed neural network.