A Bias Trick for Centered Robust Principal Component Analysis

TL;DR

This paper introduces a bias trick that automatically centers non-outliers in Robust Principal Component Analysis, leading to an optimal centering method that improves RPCA performance.

Contribution

The paper presents the first RPCA algorithm that is optimal with respect to centering by employing a novel bias trick for automatic centering.

Findings

The bias trick effectively centers non-outliers in RPCA.

The proposed algorithm achieves optimal centering performance.

Empirical results demonstrate improved robustness in RPCA.

Abstract

Outlier based Robust Principal Component Analysis (RPCA) requires centering of the non-outliers. We show a "bias trick" that automatically centers these non-outliers. Using this bias trick we obtain the first RPCA algorithm that is optimal with respect to centering.