主成分分析
数学
子空间拓扑
特征向量
稀疏PCA
降维
协方差矩阵
线性子空间
维数(图论)
模式识别(心理学)
协方差
阈值
人工智能
算法
统计
计算机科学
组合数学
量子力学
数学分析
几何学
物理
图像(数学)
出处
期刊:Annals of Statistics
[Institute of Mathematical Statistics]
日期:2013-04-01
卷期号:41 (2): 772-801
被引量:299
摘要
Principal component analysis (PCA) is a classical dimension reduction method which projects data onto the principal subspace spanned by the leading eigenvectors of the covariance matrix. However, it behaves poorly when the number of features $p$ is comparable to, or even much larger than, the sample size $n$. In this paper, we propose a new iterative thresholding approach for estimating principal subspaces in the setting where the leading eigenvectors are sparse. Under a spiked covariance model, we find that the new approach recovers the principal subspace and leading eigenvectors consistently, and even optimally, in a range of high-dimensional sparse settings. Simulated examples also demonstrate its competitive performance.
科研通智能强力驱动
Strongly Powered by AbleSci AI