On the number and the distribution of RBF centers

NeurocomputingPublished 1 March 1995

V.David Sánchez A

DOI10.1016/0925-2312(94)00074-3

Source URL

Citations15

SJR quartileQ1

SJR score1.47

SNIP1.94

TL;DR

Two observations on the design of RBF networks are reported regarding their approximation capabilities: the influence of the number and the distribution of their centers.

Abstract

Neural networks used for supervised learning need to be optimized regarding two aspects: approximation and generalization. The network needs good approximation capabilities in order to fit the training data adequately. The network also needs good generalization capabilities, because it will be used with data not seen during training. Two observations on the design of RBF networks are reported regarding their approximation capabilities: the influence of the number and the distribution of their centers. The new concept of second derivative dependent center distribution is illustrated experimentally.

Keywords

Computer ScienceEngineering

Information and ComputationDecision theoretic generalizations of the PAC model for neural net and other learning applications

871 Citations1992David Haussler

Theorems on the uniform convergence of empirical loss estimates to true expected loss rates for certain hypothesis spaces H are given, and it is shown how this implies learnability with bounded sample size, disregarding computational complexity.

Machine LearningApproximation and estimation bounds for artificial neural networks

404 Citations1994Andrew R. Barron

The analysis involves Fourier techniques for the approximation error, metric entropy considerations for the estimation error, and a calculation of the index of resolvability of minimum complexity estimation of the family of networks.

View PDF

Neural ComputationImproving the Generalization Properties of Radial Basis Function Neural Networks

328 Citations1991Chris Bishop

It is shown that a modification to the error functional allows smoothing to be introduced explicitly without significantly affecting the speed of training.

Neural NetworksOn radial basis function nets and kernel regression: Statistical consistency, convergence rates, and receptive field size

141 Citations1994Lei Xu, Adam Krzyżak +1 more

IEEE Transactions on Neural NetworksEvolving space-filling curves to distribute radial basis functions over an input space

110 Citations1994Bruce A. Whitehead, T.D. Choate

An evolutionary neural network training algorithm is proposed for radial basis function (RBF) networks that appear to have better generalization performance on the Mackey-Glass time series than corresponding networks whose centers are determined by k-means clustering.

Computers & Mathematics with ApplicationsSome extensions of radial basis functions and their applications in artificial intelligence

78 Citations1992Federico Girosi

Some extensions of the standard theory of Radial Basis Functions are shown, introduced in order to deal with a large number of examples and with problems in which different variables play very different roles.

Convergence Rates of Approximation by Translates

48 Citations1992Federico Girosi, Gabriele Anzellotti

It is shown that it is possible to define function spaces for which the rate of convergence to zero of the error is O({1 \over \sqrt n})$ in any number of dimensions.

View PDF

NeurocomputingA nonlinear network model for continuous learning

15 Citations1991Petri A. Jokinen

The dynamically capacity allocating (DCA) network model is able to learn incrementally as more information becomes available and to avoid the spatially unselective forgetting of commonly used learning algorithms.