Design of Artificial Neural Networks with a Specified Quality of Functioning

The general approach to development of methods for determination of quality of functioning of artificial neural networks any structure and destination has been formulated. This paper demonstrates complex index of quality (accuracy) of functioning of the artificial neural networks. The index considers values of functional tolerances. The article describes properties of the complex index and its use in examples of designing of artificial neural networks. These networks were used for approximating of basic mathematical functions and estimating of amplitude of harmonic signals containing noise component. Dependence of quality of functioning of artificial neural networks on a chosen training function was shown.