A neural network with competitive layers for character recognition

A structure and functioning mechanisms of a neural network with competitive layers are described. The is intended to solve the character recognition task. consists several neurons. Each layer consisting number neurons represented as layer. equal recognized classes. All have one-to-one correspondence one another input raster. every mutual lateral learning connections, which weights modified during process. There (inhibitory) relationship between all layers. This interaction realized by means “winner-take-all” (WTA) procedure aim select highest level activity.Validation has been done in experiments on handwritten digits MNIST database. demonstrated that its error rate few less than 2%, not high result, but it compensated rather fast data processing very simple mechanisms.