简介:Thispaperaddressesthehighdimensionsampleproblemindiscriminateanalysisundernonparametricandsupervisedassumptions.SincethereisakindofequivalencebetweentheprobabilisticdependencemeasureandtheBayesclassificationerrorprobability,weproposetouseaniterativealgorithmtooptimizethedimensionreductionforclassificationwithaprobabilisticapproachtoachievetheBayesclassifier.TheestimatedprobabilitiesofdifferenterrorsencounteredalongthedifferentphasesofthesystemarerealizedbytheKernelestimatewhichisadjustedinameansofthesmoothingparameter.Experimentresultssuggestthattheproposedapproachperformswell.
简介:TheproblemforcalculatingnearfieldsofEMradiationsystemsbyusingthefinitedifferencetimedomain(FD-TD)methodarediscussedandtheannularphasedarrayofdipoleantennashasbeensimulatednumericallybyuseoftheFD-TDmethod.Foratestrunthenearfieldandcurrentdistributionofthesingledipoleantennaarecalculated.Thenearfieldsoftheannularphasedarrayofdipoleantennasincentralregionfilledwithdeionizedwaterarecomputedandtheinteractionofnearfieldswithananatomically-basedinhomogeneousmodelofhumantorsoisconsideredaswell.
简介: