Abstract:Principal Component Analysis (PCA) is the most widely used unsupervised subspace learning method, and lots of its variants have been developed. With so many proposed PCA-like methods, it is still not clear that which features are better or worse for principal components, especially when the data suffers from outliers. To this end, we propose Robust Principal Component Analysis via joint ℓ 2,1 -norms minimization, which provides new insights into two crucial issues of PCA: feature selection and robustness to outliers. Unlike other PCA-like methods, the proposed method is able to select effective features for reconstruction by using the ℓ 2,1 -norm regularization term. More specific, we first use a ℓ 2,1 -norm based transformation matrix to select effective features that can effectively characterize key components (e.g., the eyes and the nose in a face image), and then use an orthogonal transformation matrix to recover the original data from the selected data representation. In this way, the key components can be well recovered by using the effective features selected by a learned transformation matrix. On the other hand, we also impose ℓ 2,1 -norm on a loss term to select clean samples to recover its same class samples but with outliers. A simple yet effective optimization algorithm is proposed to solve the resulting optimization problem. Experiments on six datasets demonstrate the effectiveness of the proposed method.

Fast Extended Inductive Robust Principal Component Analysis With Optimal Mean

Robust Principal Component Analysis Based on Maximum Correntropy Criterion

Normalized Robust PCA With Adaptive Reconstruction Error Minimization

Avoiding Optimal Mean Robust PCA/2DPCA with Non-Greedy L1-norm Maximization

Avoiding Optimal Mean ℓ 2,1 -Norm Maximization-Based Robust PCA for Reconstruction

Improved Algorithms for High-Dimensional Robust Pca

Robust principal component analysis via optimal mean by joint ℓ

Efficient Robust Principal Component Analysis Via Block Krylov Iteration and CUR Decomposition

$e^{\text{RPCA}}$: Robust Principal Component Analysis for Exponential Family Distributions

Robust Principal Component Analysis via Joint l(2,1)-Norms Minimization

Robust and Sparse Principal Component Analysis With Adaptive Loss Minimization for Feature Selection

Robust Principal Component Analysis Via Optimal Mean by Joint ℓ2,1 and Schatten P-Norms Minimization

Robust Principal Component Analysis via Discriminant Sample Weight Learning

Ensemble Principal Component Analysis

Robust Principal Component Analysis Via Joint ℓ<inf>2,1</inf>-Norms Minimization

Robust Formulation for PCA: Avoiding Mean Calculation with L2,p-norm Maximization

Robust principal component analysis: A factorization-based approach with linear complexity

Self-paced Principal Component Analysis

RES-PCA: A Scalable Approach to Recovering Low-Rank Matrices

Fast Algorithms for Robust Principal Component Analysis with an Upper Bound on the Rank.