ADAPTATION OF ARTIFICIAL NEURAL NETWORKS FOR CUDA-TECHNOLOGY APPLICATION

Abstract

The paper deals with the application of CUDA-technology on software implementation of direct and reverse passes of artificial neural network (ANN) based on back-propagation algorithm. It is shown that introduction of «imaginary» neurons helps to adapt the topology of the neural network to be trained and calculated with CUDA-technology. It is proved that «imaginary» neurons do not affect on the calculations in the back propagation algorithm.

Keywords

• cuda-технология
• «мнимый» нейрон