简介:为消除Bayes动态模型中噪声的正态性假定对模型的实用性的限制,在Bayes整体风险最小的准则下,建立了非正态Bayes动态模型状态参数向量纳向前m步Bayes最优线性预洲及其风险矩阵的循环递推方程,使正态Bayes动态模型的相关结果成为其特例。该方法可以在较大程度上拓宽Bayes动态模型及其Bayes预测的适用范围,有一定纳理论意义和实用价值。
简介:
简介:Thispaperaddressesthehighdimensionsampleproblemindiscriminateanalysisundernonparametricandsupervisedassumptions.SincethereisakindofequivalencebetweentheprobabilisticdependencemeasureandtheBayesclassificationerrorprobability,weproposetouseaniterativealgorithmtooptimizethedimensionreductionforclassificationwithaprobabilisticapproachtoachievetheBayesclassifier.TheestimatedprobabilitiesofdifferenterrorsencounteredalongthedifferentphasesofthesystemarerealizedbytheKernelestimatewhichisadjustedinameansofthesmoothingparameter.Experimentresultssuggestthattheproposedapproachperformswell.
简介:THEASYMPTOTICALLYOPTIMALEMPIRICALBAYESESTIMATIONINMULTIPLELINEARREGRESSIONMODEL¥ZHANGSHUNPU;WEILAISHENG(DepartmentofMathemati...
简介:RecentlyR.S.SinghhasstudiedtheempiricalBayes(EB)estimationinamultiplelinearregressionmodel.InthispaperweconsidertheEBtestofregressioncoefficientβforthismodel.WeworkouttheEBtestdecisionrulebyusingkernelestimationofmultivariatedensityfunctionanditsfirstorderpartialderivatives.Weobtainitsasymptoticallyoptimal(a.o.)propertyundertheconditionE||β||1<∞.ItisshownthattbeconvergenceratesofthisEBtestdecisionruleareO(n-(r-1)λ/p+r)undertheconditionE||β||pr/2-λ<∞.whereanintegerr>l,0
简介:Inthispaper,wedevotetoconstructingtheone-sidedempiricalBayes(EB)testforthelocationparameterintheGammadistributionbynonparametricmethod.Undersomemildconditions,weprovethattheEBtestisasymptoticallyoptimalwiththerateoftheorderO(n-δs/2s+1),where1/2≦δ<1ands>1isagivennaturalnumber.Anexampleisalsogiventoillustratethattheconditionsofthemaintheoremsareeasilysatisfied.
简介:文中对给定容量为n的一个伽玛分布样本,在刻度平方误差损失函数下,研究了伽玛分布参数的Bayes估计,证明了这一估计是可容许的,并给出了未知参数的Bayes区间估计.