Online Robust Subspace Tracking from Partial Information

Abstract

This paper presents GRASTA (Grassmannian Robust Adaptive Subspace Tracking Algorithm), an efficient and robust online algorithm for tracking subspaces from highly incomplete information. The algorithm uses a robust $l^1$-norm cost function in order to estimate and track non-stationary subspaces when the streaming data vectors are corrupted with outliers. We apply GRASTA to the problems of robust matrix completion and real-time separation of background from foreground in video. In this second application, we show that GRASTA performs high-quality separation of moving objects from background at exceptional speeds: In one popular benchmark video example, GRASTA achieves a rate of 57 frames per second, even when run in MATLAB on a personal laptop.

28 pages, 12 figures

References in corpus (1)

• Geometric Optimization Methods for Adaptive Filtering

Cited by in corpus (18)

• Bilinear Generalized Approximate Message Passing
• An Overview of Robust Subspace Recovery
• Online Learning with Inexact Proximal Online Gradient Descent Algorithms
• Robust PCA as Bilinear Decomposition with Outlier-Sparsity Regularization
• Memory Limited, Streaming PCA
• Iterative Grassmannian Optimization for Robust Image Alignment
• Robust Low-rank Matrix Completion via an Alternating Manifold Proximal Gradient Continuation Method
• A Survey on Matrix Completion: Perspective of Signal Processing
• Deep Neural Network Concepts for Background Subtraction: A Systematic Review and Comparative Evaluation
• Extracting low-dimensional dynamics from multiple large-scale neural population recordings by learning to predict correlations
• A Short Note on Improved ROSETA
• A Correctness Result for Online Robust PCA
• Matrix Completion with Side Information using Manifold Optimization
• Robust On-line Matrix Completion on Graphs
• Adaptive Stochastic Gradient Descent on the Grassmannian for Robust Low-Rank Subspace Recovery and Clustering
• A Manifold Approach to Learning Mutually Orthogonal Subspaces
• Online Algorithms for Factorization-Based Structure from Motion
• Online Variational Bayesian Subspace Filtering with Applications