互补约束优化的约束变尺度算法研究-应用数学专业论文.docxVIP

摘要 摘要 摘要 互补约束优化 (简称 MPCC) 是一类特殊的非线性约束最优化问题, 其约束条件 含有线性或非线性函数所构成的互补约束项. 由于其可行域的复杂性, 使得这类问题 变得难解. 本文给出了两个求解互补约束优化问题的有效算法, 具体内容如下: 第一, 提出了一个求解线性互补约束优化问题的序列线性方程组算法. 首先, 利用 非线性互补函数及特殊的磨光技术, 将原问题转化为经典的光滑非线性规划. 进而, 每 步去代的搜索方向只需通过三个系数矩阵相同的线性方程组得到. 从而, 减少了计算 量. 同时, 由于不含有任何辅助参数和人工变量, 如果算法有限步终止, 理论上可以得 到原问题的精确稳定点. 进一步, 在适当的条件下, 分析了算法具有全局收敛性和超线 性收敛速度. 第二, 针对非线性互补约束优化问题, 给出了一个变尺度型投影梯度算法. 通过摄 动技术和带扰动项的互补函数, 结合隐光滑思想将原始非线性互补约束优化问题转化 为一般约束最优化问题. 利用相应的罚函数, 将原问题进一步转化为仅含不等式约束 的非线性规划. 进而, 通过变尺度型投影技术得到显示搜索方向, 建立了一个原问题的 新算法. 在算法中, 罚参数具有自适应性, 光滑因子 μ 以变量的形式出现保证了算法 在有限步终止时, 可以得到原问题的精确稳定点. 在较弱条件下, 算法的全局收敛和超 线性收敛速度得到证明. 最后, 对上述算法进行了数值实验, 实验结果表明算法是有效的. 关键词:互补约束优化; 罚函数; 序列线性方程组; 变尺度型梯度投影; 全 局收敛; 超线性收敛速度 万方数据 – i – Abstract Abstract – – ii – 万方数据 Abstract The mathematical programs with complementarity constraints(MPCC) is a spe- cial class of nonlinear programs in which the constraints include linear or nonlinear function as complementarity constraints. Due to the complexity of the feasible region of optimization with complementarity constraints, the discussed problem become diffi- cult. In this thesis, two efficient algorithms for MPCC are presented. The achievements can be summarized into the following two aspects. We firstly proposed an sequential systems of linear equations (SSLE) algorithm for a class of optimization problems with linear complementarity constraints. By use of a special smoothing technique and nonlinear complementarity function, the original problem is equivalently transformed to a standard smoothing nonlinear optimization. At each iteration of the proposed method, the search direction is obtained by solving only three systems of linear equations with the same coefficients. In the meanwhile, our approach does not adopt any artificial variable or parameter. Thus, theoretically speaking, we conclude that the current iterate point is an exact stationary point of original problem when the algorithm stops at current point. Furthermore, the global convergence and superlinear c