Distributed Robust Principal Component Analysis

TL;DR

This paper introduces DCF-PCA, a distributed algorithm for robust principal component analysis that efficiently handles large-scale data with sparse errors, overcoming scalability issues of previous methods.

Contribution

The paper presents the first distributed RPCA algorithm based on consensus factorization, with proven convergence and applicability to large-scale problems.

Findings

DCF-PCA converges reliably in distributed settings.

The algorithm effectively handles sparse gross errors.

Scalability is improved over traditional RPCA methods.

Abstract

We study the robust principal component analysis (RPCA) problem in a distributed setting. The goal of RPCA is to find an underlying low-rank estimation for a raw data matrix when the data matrix is subject to the corruption of gross sparse errors. Previous studies have developed RPCA algorithms that provide stable solutions with fast convergence. However, these algorithms are typically hard to scale and cannot be implemented distributedly, due to the use of either SVD or large matrix multiplication. In this paper, we propose the first distributed robust principal analysis algorithm based on consensus factorization, dubbed DCF-PCA. We prove the convergence of DCF-PCA and evaluate DCF-PCA on various problem setting