Kohonen self-organizing map application to representative sample formation in the training of the multilayer perceptron
In this paper, we have considered an item of effective formation of a representative sample for training the neural network of the multilayer perceptron (MLP) type. The main problems arising in the process of the factor space division into the test, verification and training sets were formulated. An approach based on the use of clustering, that allowed one to increase the entropy of the training set was put forward. Kohonen selforganizing maps (SOM) were examined as an effective procedure of a clustering. Based on such maps, the clustering of factor spaces of different dimensions was carried out, and a representative sample was formed. To verify our approach we synthesized the MLP neural network and trained it. The training technique was performed with the sets formed both using the clustering and no doing it. The approach under consideration was concluded to have an influence on the increase in the entropy of the training set and (as a result) to lead to the quality improvement of training of MLP with the small dimensionality of the factor space.