基于Dirichlet过程混合的高斯过程模型混合采样推理  

Study on Hybrid Sampling Inference for Dirichlet Process Mixture of Gaussian Process Model

作　　者：雷菊阳[1,2] 黄克[1] 许海翔 史习智[1] 

机构地区：[1]上海交通大学机械系统与振动国家重点实验室,上海200240 [2]上海工程技术大学机械工程学院,上海201620 [3]上海特检院,上海200062

出　　处：《上海交通大学学报》2010年第2期271-275,共5页Journal of Shanghai Jiaotong University

基　　金：上海市科委基础研究资助项目(05JC14026)

摘　　要：提出了基于Dirichlet过程混合的高斯过程模型揭示复杂动态系统结构数据的多态性的内在机制.针对均值结构与协方差结构稀疏性的差异性,设计了参数先验与非参数先验来构建基于Polya urn与过松弛层采样的混合采样框架体系.该混合采样方案不但能够在统一的Metropolis-Hasting(M-H)概率评价准则下实现,而且能够最大限度地克服高斯随机走步的缺陷,方便、快速地获得马尔科夫样本链的展开.仿真结果表明,混合采样算法比高斯过程回归模型及高斯过程函数回归混合模型具有更广泛的适应性及更好的预测效果.Dirichlet process mixture of Gaussian process model was proposed to reveal the intrinsic mechanism of multi-model of complex dynamic system architecture data. As for the difference between the mean structure and covariance structure of sparsity, parametric a priori and non-parametric a priori were designed based on the hybrid sampling framework of Polya urn sampling and over-relaxed sliced sampling. The hybrid sampling will not only be implemented under the unified Metropolis-Hasting probability evaluation criteria , but also be able to overcome the shortcomings of Gaussian random walk. Markov chain samples can be quickly and easily extended. The simulation results show that the hybrid sampling algorithm has more extensive adaptabilities and more accurately predictive effect than that of Gaussian process regression model and GPFR mixture model.

关 键 词：混合采样 非参数贝叶斯推理 Dirichlet过程混合 高斯过程 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]