OPEN ACCESS Performance Comparison on Face Recognition System Using Different Variants of Back-Propagation Algorithm with Radial Basis Function Neural Networks

This paper presents the performance comparison of two architectures of neural networks: multi-layer perceptron (MLP) neural networks and radial basis function (RBF) neural networks on face recognition system (FRS). We are training MLP using different variants of back-propagation (BP) algorithm. AT&T database has been used for performance comparison. The BP is gradient descent based iterative algorithm which takes larger training time for high dimensional pattern recognition problem. Local minimum, improper learning rate and over-fitting are some of the other issues. To overcome these issues, we used RBF based FRS that is robust than other conventional methods like BP algorithm and has better performance of recognition rate. The training results show that in all the situations, RBF provides better generalization performance in compared of B