Showing papers on "Radial basis function network published in 1989"
Proceedings Article•
16 Oct 1989
TL;DR: It is concluded that nonlinear optimisation of the first layer parameters is beneficial only when a minimal network is required to solve a given problem, since the same generalisation performance can be achieved simply by using more centres and adapting only the final layer by linear optimisation.
Abstract: The author and D.S. Broomhead developed (1988) the opinion that most current feedforward layered neural networks perform a curve fitting operation in a high-dimensional space. To create the analogy, it was necessary to generalise earlier papers' assumptions, and so a mechanism for choosing radial basis functions was needed. The method involves optimisation. It is concluded that nonlinear optimisation of the first layer parameters is beneficial only when a minimal network is required to solve a given problem, since the same generalisation performance can be achieved simply by using more centres and adapting only the final layer by linear optimisation. The processing time is many orders of magnitude longer when full adaptation was used. Nonlinear optimisation cannot be used to improve the generalisation performance of the network. Choice of the nonlinearity is not crucial. >
180 citations
Book•
1 May 1989
TL;DR: A reinforcing sleeve of plastic or a similar material for the ends of hard paper reel cores or the like which form the support cores for wound goods such as paper, foils, textiles or thelike.
Abstract: A reinforcing sleeve of plastic or a similar material for the ends of hard paper reel cores or the like which form the support cores for wound goods such as paper, foils, textiles or the like. The sleeve comprises a collar adapted to rest against the end faces of the reel core and a radial, hollow projection integrated with the collar and the outer periphery of the sleeve, the projection being adapted to engage one end face of the reel core and to house a driver system. A reinforcing ring is integrated in the collar and comprises a radially extending receiving element for the driver system, the receiving element being integrated in the radial projection and reinforcing it.
174 citations
1 Jan 1989
TL;DR: The application of a radial basis functions network to a static speech pattern classification problem is described and recognition results compare well with those obtained using backpropagation and a vector-quantized hidden Markov model on the same problem.
Abstract: The application of a radial basis functions network to a static speech pattern classification problem is described. The radial basis functions network offers training times two to three orders of magnitude faster than backpropagation, when training networks of similar power and generality. Recognition results compare well with those obtained using backpropagation and a vector-quantized hidden Markov model on the same problem. A computationally efficient method of exactly solving linear networks in a noniterative fashion is also described. The method was applied to classification of vowels into 20 classes using three different types of input analysis and varying numbers of radial basis functions. The three types of input vectors consisted of linear-prediction-coding cepstral coefficient; formant tracks with frequency, amplitude, and bandwidth information; and bark-scaled formant tracks. All input analyses were supplemented with duration information. The best test results were obtained using the cepstral coefficients and 170 or more radial basis functions. >
146 citations
TL;DR: A neural network model incorporating radial basis functions is used in a speech-pattern classification problem and is compared with a back-propagation neural network models and with a vector-quantised hidden Markov model.
Abstract: A neural network model incorporating radial basis functions is used in a speech-pattern classification problem. The method is compared with a back-propagation neural network model and with a vector-quantised hidden Markov model of the same problem. Training times are over an order of magnitude faster, with similar classification results.
58 citations
TL;DR: In this paper, new constraints of radial distribution functions in the canonical ensemble of multicomponent mixtures are derived, which indicate that mixture radial distribution function are not all independent from each other.
9 citations
24 Jan 1989
TL;DR: If the network is trained (using any appropriate problem) to minimise the sum squared error over all outputs and patterns such that the output weights have minimum norm, then the output values of the trained network for any subsequent input pattern will sum to a constant.
Abstract: : Take a feed-forward layered network (such as a multilayer perceptron or a radial basis function network) which is to operate as a pattern classifier. The network may have several hidden layers, as many nodes as required and any desired nonlinearities on the hidden units. The transfer functions of the output nodes should be linear. If the network is trained (using any appropriate problem) to minimise the sum squared error over all outputs and patterns such that the output weights have minimum norm, then the output values of the trained network for any subsequent input pattern will sum to a constant. keywords: Radar, Great Britain.
2 citations
22 Nov 1989
TL;DR: A sequential adaptation scheme for neural networks is proposed that is formulated as equality constrained optimization tasks and based on the geometric point of view of pattern recognition and on methods of surface interpolation.
Abstract: A sequential adaptation scheme for neural networks is proposed. The scheme is formulated as equality constrained optimization tasks. The approach taken in developing the scheme is based on the geometric point of view of pattern recognition and on methods of surface interpolation. The set of training equations for the radial basis function network of Gaussian nodes is developed, and an experiment on its classification performance is carried out. >
1 Jan 1989
TL;DR: Two neural network models (back-propagation network and radial basis functions network) are applied to a static speech recognition problem and both models compare favourably with a vector-quantised bidden Markov model on the name problem.
Abstract: We have applied two neural network models (back-propagation network and radial basis functions network) to a static speech recognition problem. The radial basis functions network offers training times of over two orders of magnitude faster than back-propagation, when training networks to similar power and generality. We have computed recognition statistics of the two models with varying numbers of hidden units on this recognition problem. The back-propagation network may offer increased generalisation and robustness; both models compare favourably with a vector-quantised bidden Markov model on the name problem.