Kohonen self-organizing map application to representative sample formation in the training of the multilayer perceptron

In this paper, we have considered the issue of effectively forming 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, validation and training sets were formulated. An approach based on the use of clustering that allowed to increase the entropy of the training set was put forward. Kohonen's self-organizing maps (SOM) were examined as an effective clustering procedure. 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 with and without clustering. 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 MLP training with the small dimension of the factor space.