项目名称: 隐藏马尔科夫模型及粒子模型的泛函不等式与信息传输不等式
项目编号: No.11201487
项目类型: 青年科学基金项目
立项/批准年度: 2013
项目学科: 数理科学和化学
项目作者: 胡淑兰
作者单位: 中南财经政法大学
项目金额: 22万元
中文摘要: 该项目着重研究隐藏马尔科夫模型及粒子模型的统计估计的大偏差原理、泛函不等式以及信息传输不等式。这些结果是能提供有效的尾概率估计和随机过程遍历速度估计的重要工具。 (1)由于隐藏马尔科夫模型的大偏差原理、泛函不等式及信息传输不等式的相关研究非常少,本项目通过对隐藏马尔科夫模型的初始值的指数遗忘性和几何遍历性及在乘积空间的马氏过程的泛函不等式的性质,来重新寻找在隐藏马氏模型背景下,一些经典泛函不等式(Poincaréog-Sobolev等)成立的条件,使得能更一步研究隐藏马氏模型的性质。 (2)本项目通过寻找粒子模型的统计估计的泛函不等式以及信息传输不等式成立的条件,来进一步分析粒子模型的测度集中性和收敛速度的估计值。 (3)马尔科夫机制转换模型的统计估计量在实际问题中的模拟分析。
中文关键词: 大偏差原理;中偏差原理;泛函不等式;隐藏马尔可夫模型;粒子模型
英文摘要: The project aims to study large deviation principle, functional inequalities and information transportation inequalities for hidden Markov models and particle models.These results can provide some very important skills for efficiently estimating tail probability and the ergodic speed rate of stochastic processes. (1)Because the research on functional inequalities and information transportation inequalities for hidden Markov models is rare, the project is to find the conditions for functional inequalities,such as Poincare inequalities, Log-Sobolev inequalities, of hidden Markov models through the properties, like the initial exponential forgetting, geometric ergodicity and functional inequalities in product spaces for Markov processes. (2)This project is to prove and find the sufficient and necessary conditions for funcitonal inequalities and information transportation inequalities for mean field partical models and their estimators. Furthermore, we can analyse the measure concentration and the convergence speed of the statistical estimators for partical models. (3)We can analyse and simulate the Markov switching models and their statistical estimators by the real datas using some software, like R or matlab.
英文关键词: Large deviation principle;Moderate deviation principle;Functional inequality;Hidden Markov model;Particle model