The comparison of performance of mu

The comparison of performance of multilayer feed-forward perceptron (MLP) and self-organizing features maps (SOM) presents practical and scientific interest. In this paper we are providing a comparative analysis of the effectiveness of the mentioned architectures for pattern recognition of printed letters images. It should be noted that abstract features like relative placement of elements in character or junction patterns and their count are omitted in our experiments due to the fact that raster image was fed into the MLP and SOM [2] networks in the form of one-dimensional array. It was shown that MLP is more preferable for solving character recognition problem with described input data.