Specification of Training Sets and the Number of Hidden Neurons for Multilayer Perceptrons
Neural ComputationPublished 1 December 2001
L. S. Camargo, Takashi Yoneyama
Citations40
SJR quartileQ1
SJR score0.83
SNIP1.45
Generate an AI Snapshot to get a quick, structured summary of this paper.
Study Snapshot
ObjectiveStudy objective
MethodsResearch methodology
PopulationPopulation studied
Sample sizeSample sizes
OutcomesStudy outcomes here
ResultsStudy results comes here
LimitationsResearch study limitations comes here
A concise AI-generated summary of the paper will appear here once you click Generate AI Snapshot.
TL;DR
The main idea is based on the fact that Chebyshev polynomials can provide approximations to bounded functions up to a prescribed tolerance, and, in turn, a polynomial of a certain order can be fitted with a three-layer perceptron with a prescribed number of hidden neurons.
Abstract
This work concerns the selection of input-output pairs for improved training of multilayer perceptrons, in the context of approximation of univariate real functions. A criterion for the choice of the number of neurons in the hidden layer is also provided. The main idea is based on the fact that Chebyshev polynomials can provide approximations to bounded functions up to a prescribed tolerance, and, in turn, a polynomial of a certain order can be fitted with a three-layer perceptron with a prescribed number of hidden neurons. The results are applied to a sensor identification example.
Keywords
Computer ScienceEngineering
Mathematics of Control Signals and SystemsApproximation by superpositions of a sigmoidal function
13,684 Citations1989George Cybenko
Mathematics of ComputationAn Introduction to Numerical Analysis
3,101 Citations1990F. S., Kendall Atkinson
IEEE Transactions on Information TheoryUniversal approximation bounds for superpositions of a sigmoidal function
2,849 Citations1993Andrew R. Barron
The approximation rate and the parsimony of the parameterization of the networks are shown to be advantageous in high-dimensional settings and the integrated squared approximation error cannot be made smaller than order 1/n/sup 2/d/ uniformly for functions satisfying the same smoothness assumption.
Measurement Systems: Application and Design
749 Citations1966Ernest O. Doebelin
This paper aims to provide a history of ecoulement and mesures used in this discipline over a 25-year period and aims to establish a chronology of events leading up to and including the invention of EMT.
Neural NetworksUniversal Approximation Using Feedforward Neural Networks: A Survey of Some Existing Methods, and Some New Results
578 Citations1998Franco Scarselli, Ah Chung Tsoi
A unifying framework is introduced to understand existing approaches to investigate the universal approximation problem using feedforward neural networks, and two training algorithms are introduced which can determine the weights of feedforward Neural Network, with sigmoidal activation neurons, to any degree of prescribed accuracy.
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.
IEEE Transactions on Systems Man and Cybernetics Part B (Cybernetics)The Chebyshev-polynomials-based unified model neural networks for function approximation
195 Citations1998Tsu-Tian Lee, Jin-Tsong Jeng
It turns out that the Chebyshev-Polynomials-Based CPB unified model neural networks not only has the same capability of universal approximator, but also has faster learning speed than conventional feedforward/recurrent neural networks.
IEEE Transactions on Neural NetworksApproximation capability in C(R~/sup n/) by multilayer feedforward networks and related problems
194 Citations1995Tianping Chen, Hong Chen +1 more
It is found that the boundedness condition on the sigmoidal function plays an essential role in the approximation, as contrast to continuity or monotonity condition.
Neural NetworksConstructive function-approximation by three-layer artificial neural networks
77 Citations1998Shin Suzuki
Constructive theorems of three-layer artificial neural networks with trigonometric, piecewise linear, and sigmoidal hidden-layer units are proved and can easily be applied to the approximation of a non-periodic function defined in a bounded set on R(m) to R(n).
IEEE Transactions on Neural NetworksAn analytical framework for local feedforward networks
35 Citations1998Sean Weaver, Leemon C. Baird +1 more
NeurocomputingOn topology, size and generalization of non-linear feed-forward neural networks
20 Citations1997Stephan Rudolph
The correct and unique pointwise generalization capability of the new so-called similarity network topology is proved and illustrated using two examples.
Computers & Mathematics with ApplicationsA method for approximating one-dimensional functions
12 Citations1997Vasileios Basios, A.Yu. Bonushkina +1 more
A nontraditional approach to the approximation of a one-dimensional function defined on a discrete set of points is presented based on the application of an artificial feedforward neural network which realizes expansion of the function considered in orthogonal Chebyshev polynomials and which calculates the expansion coefficients during the network training.
Optimal sizing of feedforward neural networks: Case studies
10 Citations2002K.W. Lee, Henry Lam
An algorithm is proposed which can be employed to find the optimal number of hidden nodes in FNNs used for function approximation, which has advantages over the conventional trial and error method as the computational time will be reduced and there will be a lower probability of solutions getting stuck at local minima.
Neural NetworksEven on Finite Test Sets Smaller Nets may Perform Better
7 Citations1997Thomas Elsken
For feedforward multilayered neural nets, sufficient conditions are given such that for two arbitrary structures having different numbers of layers there is a finite test set S on which the optimal smaller net performs better.