Sparse bayesian learning and the relevance vector machine
The Journal of Machine Learning Research
Swarm Intelligence: Introduction and Applications
Swarm Intelligence: Introduction and Applications
International Journal of Bio-Inspired Computation
IEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics
A two-stage algorithm for identification of nonlinear dynamic systems
Automatica (Journal of IFAC)
A hybrid linear/nonlinear training algorithm for feedforward neural networks
IEEE Transactions on Neural Networks
Robust and adaptive backstepping control for nonlinear systems using RBF neural networks
IEEE Transactions on Neural Networks
Hi-index | 0.00 |
The conventional forward RBF network construction methods, such as Orthogonal Least Squares (OLS) and the Fast Recursive Algorithm (FRA), can produce a sparse network with satisfactory generalization capability. However, the RBF width, as a nonlinear parameter in the network, is not easy to determine. In the aforementioned methods, the width is always pre-determined, either by trail-and-error, or generated randomly. This will inevitably reduce the network performance, and more RBF centres may then be needed to meet a desired modelling specification. This paper investigates a new forward construction algorithm for RBF networks. It utilizes the Particle Swarm Optimization (PSO) method to search for the optimal RBF centres and their associated widths. The efficiency of this network construction procedure is retained within the forward construction scheme. Numerical analysis shows that the FRA with PSO included only needs about two thirds of the computation involved in a PSO assisted OLS algorithm. The effectiveness of the proposed technique is confirmed by a numerical simulation example.