Hybrid parallelism of artificial neural network

Abstract EN

Nowadays, artificial neural network (ANN) is considered one of the important task for solving recognition problems, such as pattern, image and speech.

This paper has designed hybrid parallelism implementation for ANN which is based on two parallel standards ; the first is multithreading standard (Open MP) for shared memory architecture at each workstation, and the second is message passing interface (MPI) for distributed memory architecture of platforms connected to the network.

We have developed an implementation of the algorithms focused on the study of hybrid parallelism.

The analytical and experimental results of the performance and effectiveness of the hybrid parallelism of ANN are presented.