%0 Journal Article
%T DOBDAlgorithmforTrainingNeuralNetwork:PartII.Application
%A WUJian-yu(吴建昱)
%A HEXiao-rong(何小荣)
%J 过程工程学报
%D 2002
%X Inthefirstpartofthearticle,anewalgorithmforpruningnetworkDynamicOptimalBrainDamage(DOBD)isintroduced.Inthispart,twocasesandanindustrialapplicationareworkedouttotestthenewalgorithm.ItisverifiedthatthealgorithmcanobtaingoodgeneralizationthroughdeletingweightparameterswithlowsensitivitiesdynamicallyandgetbetterresultthantheMarquardtalgorithmorthecross-validationmethod.Althoughtheinitialconstructionofnetworkmaybedifferent,thefinialnumberoffreeweightsprunedbytheDOBDalgorithmissimilarandthenumberisjustclosetotheoptimalnumberoffreeweights.Thealgorithmisalsohelpfultodesigntheoptimalstructureofnetwork.
%K neuralnetwork
%K DOBDalgorithm
%K Marquardtmethod
%K overfitting
%K pruning
%K training
%K application
%U http://www.jproeng.com/qikan/Cpaper/zhaiyao.asp?bsid=921