摘要
Inthispaper,aconstructivetheoryisdevelopedforapproximatingfunctionsofoneormorevariablesbysuperpositionofsigmoidalfunctions.ThisisdoneintheuniformnormaswellasintheLpnorm.Resultsforthesimultaneousapproximation,withthesameorderofaccuracy,ofafunctionanditsderivatives(whenevertheseexist),areobtained.Therelationwithneuralnetworksandradialbasisfunctionsapproximationsisdiscussed.Numericalexamplesaregivenforthepurposeofillustration.
出版日期
2013年02月12日(中国期刊网平台首次上网日期,不代表论文的发表时间)