Sequential neuron pruning algorithm for RBF network with guaranteed stability

The radial basis function (RBF) network is used in a neural network control system and the target is not only to remember the training samples but also to obtain good generalization performance. A rule of thumb for good generalization in neural systems is that the smallest system should be used to fit into the training data. Unfortunately, it is usually difficult to determine the optimal size of the RBF networks, particularly, in the sequential training procedure such as the online control problem. The proposed pruning method in this paper begins with a relatively large network, and certain units of the RBF network are dropped by examining the estimation error increment. The conic sector theory is introduced in the design of this robust neural control system, which aims at providing guaranteed boundedness for both the input-output signals and the weights of the neural network. The performance improvement of the proposed system over existing systems can be qualified in terms of better generalization ability and preventing weight shifts.