Implementation of artificial neural network on graphics processing unit for classification problems

Abstract

The artificial neural network (NN) is widely use in pattern recognition related area such as classification. After all this time, the computational process of NN is done using central processing unit (CPU). In recent years, the introduction of graphics processing unit (GPU) has opened another way to perform calculations with the advantage to speed up the calculation. In this paper, the computational process of multilayer perceptron neural network be tested on GPU using classification datasets. The performance of NN model with different number of input, hidden and output neurons are explored and compared based on the computational between GPU and CPU. The experimental result shows that the computational on GPU is much faster than CPU