|
摘要: |
本文针对带有简单凸集约束的鞍点优化问题, 提出了一个投影原始-对偶梯度方法, 算法具有对称结构且每步具有显示解.
证明了新算法的收敛性并分析了收敛速率. 最后, 泊松噪音图像恢复问题的数值结果验证了算法的有效性. |
关键词: 鞍点问题 原始-对偶 图像恢复 |
DOI: |
分类号:O224 |
基金项目:广西自然科学基金 |
|
Projection Primal-Dual Method for Saddle Point Optimization Problems with Simple Constrained |
Guo Xiao
|
Abstract: |
In this paper, we propose a projection primal-dual gradient method
for saddle point optimization problems with simple convex set constraints.
The algorithm has a symmetric structure and each step has a closed-form solution.
The convergence of the new algorithm is proved and the convergence rate is also analyzed.
The numerical results of Poisson noise image restoration problem verify the effectiveness of the new method. |
Key words: Saddle point problem Primal-dual Image restoration |