项目名称: 带加法噪声高维密度的最优小波点态估计
项目编号: No.11526150
项目类型: 专项基金项目
立项/批准年度: 2016
项目学科: 数理科学和化学
项目作者: 李蕊
作者单位: 天津理工大学
项目金额: 3万元
中文摘要: 带加法噪声密度估计是非参数统计中的重要研究方向,它在计量经济学、天文学、医学统计等领域发挥关键作用。一维小波分析已被成功地应用于解决该问题,并得到带加法噪声小波密度估计器在一维 Besov 空间中的最优收敛速度。小波在应用中取得巨大成功的主要原因在于小波具有多尺度性质,能够刻画 Besov 空间并给出密度函数的非线性(如阈值)估计。 本项目拟研究小波分析在高维 Besov 空间密度函数估计方面的应用。首先,将一维带加法噪声小波估计器推广到高维情形,并研究其在点态意义下的收敛速度;其次,给出上述小波估计的最优性分析。具体地,讨论高维 Besov 空间中密度函数与其任一估计器的风险下界。
中文关键词: 小波分析;密度估计;加法噪声;高维 Besov 空间;动态采样
英文摘要: The density estimation with additive noises is an important aspect in nonparametric statistics and plays important roles in econometrics, astronomy and medical statistics. One dimensional wavelet analysis in has been successfully applied to solve that problem, and wavelet density estimators with additive noises attain the optimal convergence rates in one dimensional Besov spaces. Wavelets have good performance in application, mainly because wavelets have important properties such as the multiresolution decomposition, characterizations for Besov spaces and provide nonlinear (e.g. thresholding) estimators for density functions. Based on those observations, this project aims to study wavelet density estimation in high dimensional Besov spaces. Firstly, we will extend one dimensional wavelet estimators to high dimensional cases and consider the convergence rates of density in pointwise sense with additive noises. Secondly, we try to give the optimality of these estimators. More precisely, lower bounds of risk between the density and any estimators will be considered in high dimensional Besov spaces.
英文关键词: wavelet analysis;density estimation;additive noise;high dimensional Besov spaces;dynamic sampling