Procedures of Tuning of Neuron net Converters of Data Presentation Form on the Basis of Programmable VLSIC

Procedures of tuning of neuron net converters of the data presentation analog form into digital code are considered. The converters have the trainable structure on the basis of programmable very large-scale integration circuits. Procedures are offered to correct weight coefficients and thresholds of neurons obtained in the result of training of the two-layer perceptron using the standard algorithm of training by example of solving a problem of the frequency conversion to a digital code.