求解凸集约束优化问题的共轭梯度的GLP投影算法

利用GLP投影技术 ,对凸约束的非线性规划问题构造了一个共轭梯度的GLP投影算法 ,在一维精确步长搜索下 ,给出了算法较强的全局收敛性结果 ,由于算法需要较小的存储量 ,特别适合于计算大规模的约束优化问题。该算法提高了梯度投影法的收敛速度。

GLP conjugate gradient projection method for solving constraint problems

The gradient projection method proposed by Calamai is modified to minimize a continuously differentiable optimization problem constrained with a closed convex set. It is shown that the modified gradient projection method has strong convergence properties using an accurate step size rule. These results have the importance to improve the recent convergence results. The method uses little storage, it is attractive for large scale problems.