DEVELOPMENT OF NEURAL NETWORK STRUCTURES OF THE FREQUENCY-CODE CONVERTER OF TWO VARIABLES AND THEIR TRAINING BASED ON THE DELTA-RULE ALGORITHM

The paper presents neural network structures of the frequency-code converter for reproducing the functional dependencies of two variables. These structures are designed for delta-rule learning. Three neural network architectures are considered, differing in the organization of layers, methods of weighing input data and a combination of signals. Issues related to training of neural network structures of converters based on modification of delta-rule algorithm are considered.