PGD算法在求解凸约束优化问题中的应用

Application of The PGD Method for Solving Convex Constrained Optimisation Problem

用拉格朗日乘子法求解约束优化问题时,涉及到求鞍点的问题,而直接求鞍点对于约束条件和变量相对较多的大问题而言,会很麻烦,这里对此情况介绍了一种新的方法,将并行变量分布算法(PGD)和Rosen的投影梯度算法(1961)结合起来使用。

When solving the constrained optimization problems using lagrangian multiplier method,it is related to saddle points.Then it is difficulty to find the saddle points when there exit a number of variables and constrain conditions.Here we introduce a new method of combining the parallel gradient distribution algorithms(PGD) and the Roses projection gradient algorithms(1961).