Matrix Completion with Prior Subspace Information via Maximizing Correlation
arXiv:2001.01152
Abstract
This paper studies the problem of completing a low-rank matrix from a few of its random entries with the aid of prior information. We suggest a strategy to incorporate prior information into the standard matrix completion procedure by maximizing the correlation between the original signal and the prior information. We also establish performance guarantees for the proposed method, which show that with suitable prior information, the proposed procedure can reduce the sample complexity of the standard matrix completion by a logarithmic factor. To illustrate the theory, we further analyze an important practical application where the prior subspace information is available. Both synthetic and real-world experiments are provided to verify the validity of the theory.
13 pages, 6 figures
References in corpus (15)
- Guaranteed Minimum-Rank Solutions of Linear Matrix Equations via Nuclear Norm Minimization
- The Augmented Lagrange Multiplier Method for Exact Recovery of Corrupted Low-Rank Matrices
- A Simpler Approach to Matrix Completion
- Modified-CS: Modifying Compressive Sensing for Problems with Partially Known Support
- Restricted strong convexity and weighted matrix completion: Optimal bounds with noise
- Direct Localization for Massive MIMO
- Semi-supervised clustering methods
- Incoherence-Optimal Matrix Completion
- Provable Inductive Matrix Completion
- Clustering Partially Observed Graphs via Convex Optimization
- Learning with the Weighted Trace-norm under Arbitrary Sampling Distributions
- Spectrally Sparse Signal Recovery via Hankel Matrix Completion with Prior Information
- Recovery of Structured Signals with Prior Information via Maximizing Correlation
- Optimal incorporation of sparsity information by weighted optimization
- Dynamic matrix recovery from incomplete observations under an exact low-rank constraint