An ART-based construction of RBF networks

Abstract

Radial basis function (RBF) networks are widely used for modeling a function from given input-output patterns. However, two difficulties are involved with traditional RBF (TRBF) networks: The initial configuration of an RBF network needs to be determined by a trial-and-error method, and the performance suffers when the desired output has abrupt changes or constant values in certain intervals. We propose a novel approach to over. come these difficulties. New kernel functions are used for hidden nodes, and the number of nodes is determined automatically by an adaptive resonance theory (ART)-like algorithm. Parameters and weights are initialized appropriately, and then tuned and adjusted by the gradient-descent method to improve the performance of the network. Experimental results have shown that the RBF networks constructed by our method have a smaller number of nodes, a faster learning speed, and a smaller approximation error than the networks produced by other methods.

From outside to hyper-globalisation: an Artificial Neural Network ordinal classifier applied to measure the extent of globalisation

“…Different types of neural networks are being used today for classification purposes, including neural networks based on a sigmoidal basis (SU), a radial basis function (RBF) (Lee and Hou 2002) and a class of multiplicative basis functions called a product unit (PU) (Martínez-Estudillo et al 2006;Schmitt 2001;Tallón-Ballesteros and Hervás-Martínez 2011). ANNs typically only use one type of basis function, but there have been some proposals to combine different functions in the hidden layer of a neural network as an alternative to traditional neural networks (Lippmann 1989), to obtain the best features of each function in the same ANN.…”

Neural Network Application for Netted Radar Information Processing Systems

“…The application of the neural network to the radar information processing in the netted radar tracking systems, bases on the conventional tracking method by means of the learning function, the large-scale parallel distributed memorizing and processing capabilities, as well as the collective operating capability of the neural network [5], accomplishes the target tracking with precision and solves the problems in the conventional method, such as the model limitation, the fusion explosion, the inference complication. At last, accomplishes the adaptive inference and improve the real-time processing ability and adaptability of the netted radar tracking systems.…”

A new e-learning achievement evaluation model based on RBF-NN and similarity filter

“…RBF-NN, one type of ANN, applies locally tuned neurons to perform function mappings, and is constructed for several different purposes such as function approximation or curve fitting and pattern recognition [5,[14][15][16][17][18][19][20]. The RBF-NN is a simple and fast learning network, and its nonlinear structures can build model for target dataset without the limitations of statistical methods (statistical distribution assumptions are necessarily made for observations).…”