Neural Network Routing Implementation using CUDA Technology

This paper presents an implementation of an algorithm for finding the routes based on neural network using the Computer Unified Device Architecture (CUDA) that exploits the parallel processing nature of neural networks. The purpose of this paper is to investigate the possibility of an efficient implementation of such algorithm and evaluate its performance (processing time to determine the route). The implementation is not affected by the network topology so it may be used in dynamic networks. The resources required are the usual ones in a personal computer equipped with CUDA graphics card. Comparing with a standard implementation using a personal computer, the obtained performance is about hundred times better using CUDA.