An Inexact Semi-smooth Newton Method on Riemannian Manifolds with Application to Duality-based Total Variation Denoising
arXiv:2102.10309 · doi:10.1137/21M1398513
Abstract
We propose a higher-order method for solving non-smooth optimization problems on manifolds. In order to obtain superlinear convergence, we apply a Riemannian Semi-smooth Newton method to a non-smooth non-linear primal-dual optimality system based on a recent extension of Fenchel duality theory to Riemannian manifolds. We also propose an inexact version of the Riemannian Semi-smooth Newton method and prove conditions for local linear and superlinear convergence that hold independent of the sign of the curvature. Numerical experiments on l2-TV-like problems with dual regularization confirm superlinear convergence on manifolds with positive and negative curvature.
References in corpus (23)
- Newton Method on Riemannian Manifolds: Covariant Alpha-Theory
- A quasi-Newton proximal splitting method
- A primal-dual hybrid gradient method for non-linear operators with applications to MRI
- A Second Order Non-Smooth Variational Model for Restoring Manifold-Valued Images
- Restoration of Manifold-Valued Images by Half-Quadratic Minimization
- Total variation regularization for manifold-valued data
- A Semismooth Newton Method for Tikhonov Functionals with Sparsity Constraints
- Second Order Differences of Cyclic Data and Applications in Variational Denoising
- A Parallel Douglas Rachford Algorithm for Minimizing ROF-like Functionals on Images with Values in Symmetric Hadamard Manifolds
- Total Generalized Variation for Manifold-valued Data
- Forward-backward truncated Newton methods for convex composite optimization
- Priors with Coupled First and Second Order Differences for Manifold-Valued Image Processing
- Mumford-Shah and Potts Regularization for Manifold-Valued Data with Applications to DTI and Q-Ball Imaging
- Exact algorithms for -TV regularization of real-valued or circle-valued signals
- A Graph Framework for Manifold-valued Data
- Convergence of inexact descent methods for nonconvex optimization on Riemannian manifolds
- Inpainting of Cyclic Data using First and Second Order Differences
- Introduction to Nonsmooth Analysis and Optimization
- First- and Second-Order Analysis for Optimization Problems with Manifold-Valued Constraints
- Fenchel Duality Theory and A Primal-Dual Algorithm on Riemannian Manifolds
- Fixed Point Algorithm Based on Quasi-Newton Method for Convex Minimization Problem with Application to Image Deblurring
- Secant Method on Riemannian Manifolds
- Damped Newton's Method on Riemannian Manifolds