Split Augmented Lagrangian Shrinkage Algorithm
Download as PPTX, PDF
2 likes1,475 views
This document proposes a Split Augmented Lagrangian Shrinkage (SALSA) algorithm to solve optimization problems involving sparsity-promoting regularization terms. It uses variable splitting to separate the data fidelity and regularization terms, and then applies an augmented Lagrangian method with soft-thresholding updates to iteratively minimize the overall cost function. Previous related approaches like iterative soft thresholding, two-step iterative soft thresholding, and fast iterative soft thresholding are also discussed.
Convert to study guideBETA
Transform any presentation into a summarized study guide, highlighting the most important points and key insights.
Split Augmented Lagrangian Shrinkage Algorithm
- 1. SPLIT AUGMENTED LAGRANGIAN SHRINKAGE ALGORITHM Mario A. T. Figueiredo, Jose M. Bioucas-Dias, and Manya V. Afonso
- 2. PROBLEM
- 3. PREVIOUS RESEARCH A number of methods are present to solve equation (1) namely Iterative Soft Thresholding (IST) Two-step IST (TwIST) Fast IST Algorithm (FISTA) Sparse Reconstruction by Separable Approximation (SpaRSA)
- 7. AL ALGORITHM
- 9. AL FOR VARIABLE SPLITTING
- 10. SALSA Soft Thresholding