Fourier-Legendre series neural networks for regression

In this study, an effective artificial neural network (ANN) using Fourier-Legendre series expansions for hidden neurons are successfully proposed, named FLSNN. The performance of the FLSNN was examined for regulation problems in machine learning applications. The results of two numerical examples of Boston price regulation problems using FLSNN are compared with classical well-known two ANNs, i.e. the backward propagation neural networks (BPANN) and radial basis function neural networks (RBFNN). It is shown that FLS offers high R-square in regression analyses.