Noisy Matrix Completion under Sparse Factor Models
arXiv:1411.0282 · doi:10.1109/TIT.2016.2549040
Abstract
This paper examines a general class of noisy matrix completion tasks where the goal is to estimate a matrix from observations obtained at a subset of its entries, each of which is subject to random noise or corruption. Our specific focus is on settings where the matrix to be estimated is well-approximated by a product of two (a priori unknown) matrices, one of which is sparse. Such structural models - referred to here as "sparse factor models" - have been widely used, for example, in subspace clustering applications, as well as in contemporary sparse modeling and dictionary learning tasks. Our main theoretical contributions are estimation error bounds for sparsity-regularized maximum likelihood estimators for problems of this form, which are applicable to a number of different observation noise or corruption models. Several specific implications are examined, including scenarios where observations are corrupted by additive Gaussian noise or additive heavier-tailed (Laplace) noise, Poisson-distributed observations, and highly-quantized (e.g., one-bit) observations. We also propose a simple algorithmic approach based on the alternating direction method of multipliers for these tasks, and provide experimental evidence to support our error analyses.
arXiv comment42 Pages, 7 Figures, Submitted to IEEE Transactions on Information Theory
References in corpus (20)
- The Convex Geometry of Linear Inverse Problems
- A Simpler Approach to Matrix Completion
- Restricted strong convexity and weighted matrix completion: Optimal bounds with noise
- A geometric analysis of subspace clustering with outliers
- Matrix Completion from Noisy Entries
- Robust subspace clustering
- A convex model for non-negative matrix factorization and dimensionality reduction on physical space
- Compressed sensing performance bounds under Poisson noise
- Factoring nonnegative matrices with linear programs
- Exact Recovery of Sparsely-Used Dictionaries
- On the Identifiability of Overcomplete Dictionaries via the Minimisation Principle Underlying K-SVD
- Multiscale likelihood analysis and complexity penalized estimation
- High-dimensional estimation with geometric constraints
- High-Rank Matrix Completion and Subspace Clustering with Missing Data
- On the Local Correctness of L^1 Minimization for Dictionary Learning
- Local stability and robustness of sparse dictionary learning in the presence of noise
- Probabilistic low-rank matrix completion on finite alphabets
- Iterative Hard Thresholding Methods for Regularized Convex Cone Programming
- Computing a Nonnegative Matrix Factorization -- Provably
- Sparse Factor Analysis for Learning and Content Analytics
Cited by in corpus (11)
- An overview of low-rank matrix recovery from incomplete observations
- Poisson Matrix Recovery and Completion
- A Novel Approach to Quantized Matrix Completion Using Huber Loss Measure
- Low-rank matrix completion and denoising under Poisson noise
- Learning Latent Features with Pairwise Penalties in Low-Rank Matrix Completion
- Noisy Inductive Matrix Completion Under Sparse Factor Models
- Poisson Matrix Completion
- Categorical Matrix Completion
- On Learning Sparsely Used Dictionaries from Incomplete Samples
- Noisy Tensor Completion for Tensors with a Sparse Canonical Polyadic Factor
- Reconstruction Error Bounds for Compressed Sensing under Poisson Noise using the Square Root of the Jensen-Shannon Divergence