| 摘要: |
| 考虑目标函数能够分解成n个独立的凸函数,其约束条件为线性约束的可分凸优化问题.呈现了一种推广的预测矫正邻近乘子法来求解可分凸优化问题.算法在迭代中利用二次项代替了增广拉格朗日函数的增广项，算法既有邻近乘子法的特性，又有可以平行计算，并且在较弱的条件下，能保证全局收敛. |
| 关键词: 可分凸优化 增广拉格朗日 邻近法 分裂法 |
| DOI： |
| 分类号: |
| 基金项目: |
|
| Separable Convex Optimization Problem Is Solved by Promoted Pre correction Neighboring Point Method |
|
LUO Li
|
| Abstract: |
| Objective function can be decomposed into n independent convex functions, and the constraint conditions for linear constraint can be divided into convex optimization problem. This paper presents a kind of promotion forecast correction multiplier method to solve separable convex optimization problem, this algorithm uses the second item in the iteration to replace the augmented Lagrangian function of the augmented items, and the proposed algorithm has both the characterization of neighboring multiplier method and parallel computation as well as can guarantee the global convergence under weak condition. |
| Key words: separable convex optimization augmented Lagrangian neighboring method disintegrating method |
