郑州大学学报（理学版）2019，Vol.51Issue(1)：39-43,5.DOI:10.13705/j.issn.1671-6841.2017336
条件区间分位数超高维特征筛选研究
Feature Screening for Ultrahigh Dimensional Data Based on the Conditional Interval Quantile
摘要
Abstract
Feature screening was an important step for model dimension reduction of ultrahigh dimension-al data. Focusing on this problem, to tackle the instability of the feature screening procedure based on the conditional quantile technique when the given quantile values had small disturbance, one global quantile technique was introduced. The generalized feature screening procedure based on the conditional interval quantile was proposed. The theoretical proof and numerical simulations were completed to prove the pro-posed screening procedure could processe the sure screening property and, showed its finite sample prop-erties.关键词
超高维/特征筛选/区间分位数/确定性筛选性质Key words
ultrahigh dimension/feature screening/interval quantile/sure screening property分类
数理科学引用本文复制引用
来鹏,张洁,季静雯..条件区间分位数超高维特征筛选研究[J].郑州大学学报（理学版）,2019,51(1):39-43,5.基金项目
国家自然科学基金项目(11771215) （11771215）
江苏省自然科学基金项目(BK20161530) （BK20161530）
江苏省"青蓝工程"项目(2016) （2016）
国家社科基金重大项目(16ZDA047). （16ZDA047）
