ADMiRA: Atomic Decomposition for Minimum Rank Approximation
From MaRDI portal
Publication:5281301
Abstract: We address the inverse problem that arises in compressed sensing of a low-rank matrix. Our approach is to pose the inverse problem as an approximation problem with a specified target rank of the solution. A simple search over the target rank then provides the minimum rank solution satisfying a prescribed data approximation bound. We propose an atomic decomposition that provides an analogy between parsimonious representations of a sparse vector and a low-rank matrix. Efficient greedy algorithms to solve the inverse problem for the vector case are extended to the matrix case through this atomic decomposition. In particular, we propose an efficient and guaranteed algorithm named ADMiRA that extends CoSaMP, its analogue for the vector case. The performance guarantee is given in terms of the rank-restricted isometry property and bounds both the number of iterations and the error in the approximate solution for the general case where the solution is approximately low-rank and the measurements are noisy. With a sparse measurement operator such as the one arising in the matrix completion problem, the computation in ADMiRA is linear in the number of measurements. The numerical experiments for the matrix completion problem show that, although the measurement operator in this case does not satisfy the rank-restricted isometry property, ADMiRA is a competitive algorithm for matrix completion.
Cited in
(36)- Finding low-rank solutions via nonconvex matrix factorization, efficiently and provably
- Rank-constrained optimization and its applications
- Low-rank matrix recovery with Ky Fan 2-\(k\)-norm
- Penalty decomposition methods for rank minimization
- Low-rank dynamic mode decomposition: an exact and tractable solution
- Decomposition into low-rank plus additive matrices for background/foreground separation: a review for a comparative evaluation with a large-scale dataset
- An adaptation for iterative structured matrix completion
- A relaxed interior point method for low-rank semidefinite programming problems with applications to matrix completion
- Smooth singular value thresholding algorithm for low-rank matrix completion problem
- Fixed-rank matrix factorizations and Riemannian low-rank optimization
- CGIHT: conjugate gradient iterative hard thresholding for compressed sensing and matrix completion
- Orthogonal rank-one matrix pursuit for low rank matrix completion
- Enhancing matrix completion using a modified second-order total variation
- Matrix rigidity and the ill-posedness of robust PCA and matrix completion
- Exact minimum rank approximation via Schatten \(p\)-norm minimization
- Learning non-parametric basis independent models from point queries via low-rank methods
- Low rank matrix recovery from rank one measurements
- Convergence of projected Landweber iteration for matrix rank minimization
- Uniqueness conditions for low-rank matrix recovery
- Low-rank matrix completion via preconditioned optimization on the Grassmann manifold
- Matrix recipes for hard thresholding methods
- An alternating minimization method for matrix completion problems
- Stable low-rank matrix recovery via null space properties
- ADMiRA
- Convergence analysis of projected gradient descent for Schatten-\(p\) nonconvex matrix recovery
- A penalty decomposition method for rank minimization problem with affine constraints
- Recovery of simultaneous low rank and two-way sparse coefficient matrices, a nonconvex approach
- Guarantees of Riemannian optimization for low rank matrix recovery
- Convergence of fixed-point continuation algorithms for matrix rank minimization
- Guarantees of Riemannian optimization for low rank matrix completion
- Implicit regularization in nonconvex statistical estimation: gradient descent converges linearly for phase retrieval, matrix completion, and blind deconvolution
- Generalizing CoSaMP to signals from a union of low dimensional linear subspaces
- A non-convex algorithm framework based on DC programming and DCA for matrix completion
- Solving a low-rank factorization model for matrix completion by a nonlinear successive over-relaxation algorithm
- Matrix completion under interval uncertainty
- Minimum \( n\)-rank approximation via iterative hard thresholding
This page was built for publication: ADMiRA: Atomic Decomposition for Minimum Rank Approximation
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5281301)
