基于MCMC的混合α稳定分布参数贝叶斯推理  

Bayesian Inference on Parameters for Mixtures of α-Stable Distributions Based on Markov Chain Monte Carlo

作　　者：陈亚军[1] 刘丁[1] 梁军利[1] 

机构地区：[1]西安理工大学自动化与信息工程学院,陕西西安710048

出　　处：《西安理工大学学报》2012年第4期385-391,共7页Journal of Xi'an University of Technology

基　　金：国家科技重大专项资助项目(2009X02011001);国家自然科学基金资助项目(61075044;61172123)

摘　　要：为解决非高斯信号较难描述这一难点问题,提出一种基于马尔科夫链蒙特卡罗方法的混合α稳定分布参数的贝叶斯推理方法。构建了混合稳定分布分层的贝叶斯图模型,利用Gibbs抽样实现了混合权值和分配参数z的估计,基于Metropolis算法完成了每个分布元中4个参数的估计。仿真结果表明,该方法能够准确地估计出混合α稳定分布中的各个参数,具有很好的鲁棒性和灵活性,可用于对非高斯信号或数据进行建模。To solve the difficult problem of non-Gaussian signal difficult to be described,this paper suggests a method of Bayesian inference on parameter for mixtures of α-stable distributions based on Markov Chain Monte Carlo.The hierarchical Bayesian graph model is constructed.Gibbs sampling algorithm is used to achieve the estimation of the mixing weights and allocation parameter z.The 4 parameter estimations in each distribution component are completed on the basis of Metropolis algorithm.The simulation results show that the method can accurately estimate the parameters of mixture of α-stable distributions,and it has good robustness and flexibility,whereby the method can be used to establish the model for non-Gaussian signal or data.

关 键 词：混合α稳定分布 马尔可夫链蒙特卡罗 Metropolis算法 GIBBS抽样 非高斯信号 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]