Multi-objective optimization of radial basis function neural network training using genetic algorithm

Radial basis function neural network (RBFNN) is an artificial feedforward neural network that uses radial basis functions as activation functions in the hidden layer. The output of the RBFNN is a linear combination of the outputs of the hidden layer. This paper presents a multi-objective model of radial basis function neural networks for training. This model aims to satisfy two objectives: the first is the sum of all distances between the input vector and the corresponding center for the selected neurons in the hidden layer and the second is the overall error of the RBFNN, which is defined as the error between the computed output and the expected output. To solve this model, we will use an approach based on genetic algorithms, which allows us to determine an appropriate partitioning of the input data and the optimal weight of the output layer that gives us a good generalization. Numerical results show the performance of the theoretical results presented in this paper, as well as the advantages of the new modeling.