In this article an architecture and learning algorithm for the modular neural network, where the hidden layer is formed by a general regression and radial basis function neural networks, have been proposed. These networks are parallel connected to the input layer and trained independently, following which the optimization with respect to the network output accuracy is performed. Formed by neural network models based on memory and optimization, the proposed modular neural network provides a high accuracy both in early learning stages and when data set could grow in real time.

  Повний текст PDF - 298.154 Kb    Зміст випуску     Цитування публікації

  Якщо, ви не знайшли інформацію про автора(ів) публікації, маєте бажання виправити або відобразити більш докладну інформацію про науковців України запрошуємо заповнити "Анкету науковця"