Universal approximation using radial-basis-function networks
Neural Computation
Approximation and radial-basis-function networks
Neural Computation
Handbook of Neural Network Signal Processing
Handbook of Neural Network Signal Processing
Applications of Neural Networks in Electromagnetics
Applications of Neural Networks in Electromagnetics
Static and Dynamic Neural Networks: From Fundamentals to Advanced Theory
Static and Dynamic Neural Networks: From Fundamentals to Advanced Theory
Null steering beamformer based on RBF neural networks
ICCOM'08 Proceedings of the 12th WSEAS international conference on Communications
Enhanced metamodeling techniques for high-dimensional IC design estimation problems
Proceedings of the Conference on Design, Automation and Test in Europe
Hi-index | 0.00 |
This paper considers the problem of reliability of radial basis function neural network (RBFNN) based smart antenna, applied for direction of arrival estimation. One of the main parameters in these networks is the number of neurons in the hidden layer. Once this parameter is chosen another important issue is at which level we can count on a proper functionality of the hidden layer, which is mainly determined by a proper functionality of its neurons. After the presentation of the architecture and the basic concept of the RBFNN, we will discuss the problem of the hidden layer dimension. This will provide necessary information for the investigation of the reliability of these networks. Namely, we will assume a failure of some percentage of the neurons in the hidden layer and we will present the resulting mean error increase. The results of computer simulations will show us that the networks with less neurons in the hidden layer are more reliable. But when we need more precise direction of arrival estimations we should use more neurons in the hidden layer and more training samples. In this case the neurons in the hidden layer must be with long time of a proper functionality, since the mean error increase is much larger if some of the neurons fail.