【摘要】 本文讨论了核为指数型分布的混合的识别问题,给出一个主要定理来确定可识别的充分条件,以及刻划先验分布的方法,并应用于可靠性的贝叶斯方法中。更多还原

【Abstract】 In this paper We consider the identifiability of mixtures in which the kernel belo- ngs to a family of exponential type distributions. A main theorem shows how to determine the mixing distribution of the mixture, i. e., for the donsity f（x, θ） C（θ）K （x）e^θx, （x>0）, and the mixture h（x）=（?）f（x, θ）g（θ）dθ. If integral from n=-∞ to ∞（|（h（z））/（K（z）））|dt<∞, z=x+it,x∈R⁺,t∈R,) the mixture is identifiable, and mixing g（θ） is uniquely determined by g（θ）=1/（2πiC（θ））integral from n=σ-iθ to σ-iθ(h（z）/K（z）e^-2θdz z=x+it,σ>0)更多还原