Title :
Synthesis of feedforward networks in supremum error bound
Author :
Ciesielski, Krzysztof ; Sacha, Jaroslaw P. ; Cios, Krzysztof J.
Author_Institution :
Dept. of Math., West Virginia Univ., Morgantown, WV, USA
fDate :
11/1/2000 12:00:00 AM
Abstract :
The main result of this paper is a constructive proof of a formula for the upper bound of the approximation error in L∞ (supremum norm) of multidimensional functions by feedforward networks with one hidden layer of sigmoidal units and a linear output. This result is applied to formulate a new method of neural-network synthesis. The result can also be used to estimate complexity of the maximum-error network and/or to initialize that network´s weights. An example of the network synthesis is given.
Keywords :
computational complexity; feedforward neural nets; multilayer perceptrons; L∞ norm; approximation error upper bound; complexity estimation; feedforward network synthesis; linear output; maximum-error network; multidimensional functions; network weight initialization; neural-network synthesis; sigmoidal units; supremum error bound; supremum norm; Approximation error; Artificial neural networks; Indium tin oxide; Intelligent networks; Minimization methods; Multidimensional systems; Network synthesis; Network topology; Neurons; Upper bound;
Journal_Title :
Neural Networks, IEEE Transactions on
