Det-CGD: Compressed Gradient Descent with Matrix Stepsizes for Non-Convex Optimization.
Hanmin LiAvetik G. KaragulyanPeter RichtárikPublished in: ICLR (2024)
Keyphrases
- convex optimization
- low rank
- low rank and sparse
- operator splitting
- low rank matrix
- interior point methods
- convex optimization problems
- trace norm
- convex relaxation
- total variation
- primal dual
- cost function
- norm minimization
- singular values
- loss function
- matrix completion
- semidefinite
- basis pursuit
- kernel matrix
- convex formulation
- data matrix
- augmented lagrangian
- singular value decomposition
- denoising
- higher order
- computer vision
- image denoising
- objective function
- image segmentation