High Dimensional Low Rank plus Sparse Matrix Decomposition

Abstract

This paper is concerned with the problem of low rank plus sparse matrix decomposition for big data. Conventional algorithms for matrix decomposition use the entire data to extract the low-rank and sparse components, and are based on optimization problems with complexity that scales with the dimension of the data, which limits their scalability. Furthermore, existing randomized approaches mostly rely on uniform random sampling, which is quite inefficient for many real world data matrices that exhibit additional structures (e.g. clustering). In this paper, a scalable subspace-pursuit approach that transforms the decomposition problem to a subspace learning problem is proposed. The decomposition is carried out using a small data sketch formed from sampled columns/rows. Even when the data is sampled uniformly at random, it is shown that the sufficient number of sampled columns/rows is roughly O(rμ), where μis the coherency parameter and r the rank of the low rank component. In addition, adaptive sampling algorithms are proposed to address the problem of column/row sampling from structured data. We provide an analysis of the proposed method with adaptive sampling and show that adaptive sampling makes the required number of sampled columns/rows invariant to the distribution of the data. The proposed approach is amenable to online implementation and an online scheme is proposed.

IEEE Transactions on Signal Processing

References in corpus (5)

• Sparsity and Incoherence in Compressive Sampling
• Subspace Learning and Imputation for Streaming Big Data Matrices and Tensors
• Randomized Robust Subspace Recovery for High Dimensional Data Matrices
• Identifying Outliers in Large Matrices via Randomized Adaptive Compressive Sampling
• Recovery of Coherent Data via Low-Rank Dictionary Pursuit

Cited by in corpus (16)

• Coherence Pursuit: Fast, Simple, and Robust Principal Component Analysis
• Subspace-Orbit Randomized Decomposition for Low-rank Matrix Approximation
• Screen Content Image Segmentation Using Sparse-Smooth Decomposition
• Robust PCA by Manifold Optimization
• Fast Algorithms for Demixing Sparse Signals from Nonlinear Observations
• Evolutionary Self-Expressive Models for Subspace Clustering
• Scalable and Robust Community Detection with Randomized Sketching
• Spatial Random Sampling: A Structure-Preserving Data Sketching Tool
• Low Rank Matrix Recovery with Simultaneous Presence of Outliers and Sparse Corruption
• Robust and Scalable Column/Row Sampling from Corrupted Big Data
• Efficient Neural Network Approximation of Robust PCA for Automated Analysis of Calcium Imaging Data
• Compressive PCA for Low-Rank Matrices on Graphs
• Leveraging Subspace Information for Low-Rank Matrix Reconstruction
• Multi-View Task-Driven Recognition in Visual Sensor Networks
• Enhanced image approximation using shifted rank-1 reconstruction
• Compressed Randomized UTV Decompositions for Low-Rank Approximations and Big Data Applications