Robust Partial Least Squares Using Low Rank and Sparse Decomposition

TL;DR

This paper introduces a robust partial least squares framework that employs low-rank and sparse decomposition to improve regression and dimensionality reduction in data contaminated with outliers.

Contribution

It presents a novel method combining low-rank and sparse matrix decomposition with PLS for robust regression and dimensionality reduction amidst outliers.

Findings

Effective in handling outliers during regression

Improves estimation of low-dimensional data manifolds

Demonstrates superior performance in experiments

Abstract

This paper proposes a framework for simultaneous dimensionality reduction and regression in the presence of outliers in data by applying low-rank and sparse matrix decomposition. For multivariate data corrupted with outliers, it is generally hard to estimate the true low dimensional manifold from corrupted data. The objective of the proposed framework is to find a robust estimate of the low dimensional space of data to reliably perform regression. The effectiveness of the proposed algorithm is demonstrated experimentally for simultaneous regression and dimensionality reduction in the presence of outliers in data.