The article is devoted to the development of sustainable and robust algorithms for setting radial basis network parameters (RBS). RBS is used for a wide range of identification, control, signal and image processing tasks. Instead the most widely used method of least squares for training the neural network it is proposed to use single-step algorithms that combine the properties of the basic algorithms and stochastic approximation algorithms. This approach helps to significantly accelerate the learning process. We give simulation results that confirm the effectiveness of the proposed algorithms for settings parameters of the basic functions.
neuron network, base function, one-step algorithm, radially-base network