چکیده:
This paper extends the sequential learning algorithm strategy of two different types of adaptive radial basis function-based (RBF) neural networks, i.e. growing and pruning radial basis funcuon (GAP-RBF) and minimal resource allocation network (MRAN) to cater for on-line identification of non-linear systems. The original sequential learning algorithm is based on the repetitive utilization of sequential input-output data in order to accomplish the training phase.Some interesting modifications have been proposed in the growing and pruning neurons criteria of the original GAP-RBF neural network to make the resulting modified GAP-RB P (MGAP-RBF) neural network suitable for on-1 inc system identification applications. The Unscented Kalman Filter (UKF) has been proposed as a new learning algorithm to update the parameters of MRAN, GAP-RBr «»o MGAP-RBF neural networks. Moreover, to keep the resulting parameter estimation routines more sensitivefactor strategy has been included in the UKF learning algorithm. Thecontinuous stirred tank reactor (CSTR) and the chaotic Mackey Glass time- series as two different benchmark problems. The resulting performances of the MRAN, GAP-RB F and the proposed MGAP-RBF neural networks being estimated with the extended Kalman filter (EKF) or the U Kf learning algorithm have been evaluated for comparison purposes. Simulation results show the superiority of the proposed MGAP-RBP neural network estimated with the UKF learning algorithm.
خلاصه ماشینی:
e. growing and pruning radial basis funcuon (GAP-RBF) and minimal resource allocation network (MRAN) to cater for on-line identification of non-linear systems.
The Unscented Kalman Filter (UKF) has been proposed as a new learning algorithm to update the parameters of MRAN, GAP-RBr «»o MGAP-RBF neural networks.
The resulting performances of the MRAN, GAP-RB F and the proposed MGAP-RBF neural networks being estimated with the extended Kalman filter (EKF) or the U Kf learning algorithm have been evaluated for comparison purposes.
These models are used for simulations, predictions, analysis of the system’s behavior, design of model-based controllers, and so forth Radial Basis Function (RBF) networks have been popularly used in many applications in recent times due to their ability to approximate complex non-linear mappings directly front the input-output data with a simple topological structure and ease of implementation of dynamic and adaptive network architectures.
[ 1] proposed a simple sequential growing and pruning algorithm based on the relationship between the significance of a neuron and the required learning accuracy for‘ RBF networks, referred to is GAP-RBF.
steriori estimate of the error covariance is computed as: (View the image of this page) THE MODIFIED GAP-RBF NEURAL NETWORK AND THE UKF LEARNING ALGORITHM Some desired modifications have been proposed in the growing and pruning neurons criteria to make the resulting MGAP-RBF neural network suitable for on-line system identification applications.
The resulting performances are compared with the original GAP-RBF und MRAN neural networks being estimated with both the EKF and UKF learning algorithms.
(
, 
, 
)
