Comparative study of MLP and RBF algorithms for data classification

This paper compare the performance of various multilayer perceptron (MLP) and radial basis function (RBF) neural networks on classification problems using Matlab’s Neural network toolbox. We tested the studied networks on data sets such as iris dataset, cancer dataset, glass dataset, thyroid dataset and etc. Several evaluation parameters, such as number of mean square of error (MSE), number of neurons in hidden layer, training time and accuracy (ACC), were taken into account during performance comparison of the algorithms. The results show that the Levenberg-Marquardt training algorithm and Radial basis function neural network is frequently faster and achieves better accuracy than the other algorithms for moderate size problems.

Keywords:

Classification , Back propagation , multilayer perceptron , Conjugate Gradient , Quasi-Newton , Levenberg-Marquardt , Radial basis function.