Abstract:We propose a theory for matrix completion that goes beyond the low-rank structure commonly considered in the literature and applies to general matrices of low description complexity. Specifically, complexity of the sets of matrices encompassed by the theory is measured in terms of Hausdorff and upper Minkowski dimensions. Our goal is the characterization of the number of linear measurements, with an emphasis on rank-$1$ measurements, needed for the existence of an algorithm that yields reconstruction, either perfect, with probability 1, or with arbitrarily small probability of error, depending on the setup. Concretely, we show that matrices taken from a set $\mathcal{U}$ such that $\mathcal{U}-\mathcal{U}$ has Hausdorff dimension $s$ can be recovered from $k>s$ measurements, and random matrices supported on a set $\mathcal{U}$ of Hausdorff dimension $s$ can be recovered with probability 1 from $k>s$ measurements. What is more, we establish the existence of recovery mappings that are robust against additive perturbations or noise in the measurements. Concretely, we show that there are $\beta$-Hölder continuous mappings recovering matrices taken from a set of upper Minkowski dimension $s$ from $k>2s/(1-\beta)$ measurements and, with arbitrarily small probability of error, random matrices supported on a set of upper Minkowski dimension $s$ from $k>s/(1-\beta)$ measurements. The numerous concrete examples we consider include low-rank matrices, sparse matrices, QR decompositions with sparse R-components, and matrices of fractal nature.

Companion unit lower Hessenberg matrices

Condition numbers of Hessenberg companion matrices

Bilinear characterizations of companion matrices

Nil-clean companion matrices

On Generalized Schur Complement of Nonstrictly Diagonally Dominant Matrices and General H-matrices

The companion section for classical groups

Group-Like algebras and Hadamard matrices

Companion matrix, Vandermonde matrix, Jordan form, Interpolating Polynomials, and Linear Transformations

Perron Complements Of Diagonally Dominant Matrices And H-Matrices

On certain matrix algebras related to quasi-Toeplitz matrices

Smith forms of matrices in Companion Rings, with group theoretic and topological applications

On Companion sequences associated with Leonardo quaternions: Applications over finite fields

On The Perron Complements Of H-Matrices

Hadamard matrices in centraliser algebras of monomial representations

An upper $J$- Hessenberg reduction of a matrix through symplectic Householder transformations

Completion of Matrices with Low Description Complexity

Uniform $\mathcal{H}$-matrix Compression with Applications to Boundary Integral Equations

General H-matrices and Their Schur Complements

Logarithmically Sparse Symmetric Matrices

Diagonalizable higher degree forms and symmetric tensors