On the Convergence of the Iterative Shrinkage/Thresholding Algorithm With a Weakly Convex Penalty

Abstract

We consider the iterative shrinkage/thresholding algorithm (ISTA) applied to a cost function composed of a data fidelity term and a penalty term. The penalty is non-convex but the concavity of the penalty is accounted for by the data fidelity term so that the overall cost function is convex. We provide a generalization of the convergence result for ISTA viewed as a forward-backward splitting algorithm. We also demonstrate experimentally that for the current setup, using large stepsizes in ISTA can accelerate convergence more than existing schemes proposed for the convex case, like TwIST or FISTA.

References in corpus (1)

• Wavelet methods in statistics: Some recent developments and their applications

Cited by in corpus (17)

• Sparse Regularization via Convex Analysis
• Total Variation Denoising via the Moreau Envelope
• Enhanced Low-Rank Matrix Approximation
• Enhanced Sparsity by Non-Separable Regularization
• Linearly Involved Generalized Moreau Enhanced Models and Their Proximal Splitting Algorithm under Overall Convexity Condition
• A Wavelet Based Sparse Row-Action Method for Image Reconstruction in Magnetic Particle Imaging
• Variable Smoothing for Weakly Convex Composite Functions
• Nonconvex Optimization via MM Algorithms: Convergence Theory
• Monotone Lipschitz-Gradient Denoiser: Explainability of Operator Regularization Approaches Free From Lipschitz Constant Control
• The Douglas-Rachford Algorithm for Weakly Convex Penalties
• A Convex-Nonconvex Strategy for Grouped Variable Selection
• Sparsity Within and Across Overlapping Groups
• Continuous Relaxation of Discontinuous Shrinkage Operator: Proximal Inclusion and Conversion
• Non-convex Lasso-kind approach to compressed sensing for finite-valued signals
• On the convergence rate improvement of a splitting method for finding the resolvent of the sum of maximal monotone operators
• A Convex-Nonconvex Framework for Enhancing Minimization Induced Penalties
• Emission tomography with a multi-bang assumption on attenuation