Abstract:We study the matrix completion problem: an underlying matrix P is low rank, with incoherent singular vectors, and a random matrix A is equal to P on a (uniformly) random subset of entries of size dn . All other entries of A are equal to zero. The goal is to retrieve information on P from the observation of A . Let be the random matrix where each entry of A is multiplied by an independent -Bernoulli random variable with parameter 1/2. This paper is about when, how and why the non-Hermitian eigen-spectra of the matrices and captures more of the relevant information about the principal component structure of A than the eigen-spectra of and . We show that the eigenvalues of the asymmetric matrices and with modulus greater than a detection threshold are asymptotically equal to the eigenvalues of and and that the associated eigenvectors are aligned as well. The central surprise is that by intentionally inducing asymmetry and additional randomness via the matrix, we can extract more information than if we had worked with the singular value decomposition (SVD) of A . The associated detection threshold is asymptotically exact and is non-universal since it explicitly depends on the element-wise distribution of the underlying matrix P . We show that reliable, statistically optimal but not perfect matrix recovery, via a universal data-driven algorithm, is possible above this detection threshold using the information extracted from the asymmetric eigen-decompositions. Averaging the left and right eigenvectors provably improves estimation accuracy but not the detection threshold. Our results encompass the very sparse regime where d is of order 1 where matrix completion via the SVD of A fails or produces unreliable recovery. We define another variant of this asymmetric principal component analysis procedure that bypasses the randomization step and has a detection threshold that is smaller by a constant factor but with a computational cost that is larger by a polynomial factor of the number of observed entries. Both detection thresholds allow to go beyond the barrier due to the well-known information theoretical limit for exact matrix completion found in the literature.

Completion Problems and Sparsity for Kemeny's Constant

On Kemeny's constant and stochastic complement

Completion of Matrices with Low Description Complexity

A Note on Sparse Solutions of Sparse Linear Systems

Quasi-Hankel low-rank matrix completion: a convex relaxation

Low rank matrix completion by alternating steepest descent methods

The Sparse Solution to $\mathcal{KS}$-Tensor Complementarity Problems

Matrix Completion and Low-Rank SVD via Fast Alternating Least Squares

Missing Entries Matrix Approximation and Completion

The CP-Matrix Completion Problem.

A non-backtracking method for long matrix and tensor completion

Optimal spectral norm rates for noisy low-rank matrix completion

A Fast Spectral Relaxation Approach to Matrix Completion Via Kronecker Products.

Approximate Method of Variational Bayesian Matrix Factorization/Completion with Sparse Prior

Detection Thresholds in Very Sparse Matrix Completion

The Sparse Solution to KS-tensor Complementarity Problems

On the Null Space Constant for ${\ell _p}$ Minimization

Optimal Low-Rank Matrix Completion: Semidefinite Relaxations and Eigenvector Disjunctions

Matrix Completion for the Independence Model

Matrix Completion with Sparse Noisy Rows

Near-Optimal Time-Sparsity Trade-Offs for Solving Noisy Linear Equations