凸规划宽邻域原始-对偶势下落内点算法及其扩展.pdfVIP

内 容 摘 要 线性规划与非线性规划是最优化理论的两个重要分支，许多实际问题可以抽象 成这两类问题．线性规划是一类特殊的凸规划，而凸规划又是一类简单的非线性规 划．可见，凸规划与优化中的两大类规划有着密切联系．因此，对它的研究无论从 理论价值还是实际应用方面来讲都是很有意义的．凸二次规划是一类简单的凸规划 问题，研究它的算法不仅仅是为了解决凸二次规划本身，更是为了更好地求解其他 非线性规划问题，因为一般的非线性函数在其极小点附近可以用一个二次函数很好 地近似． 本文着重研究了凸二次规划的几个宽邻域内点算法，并分析了它们的多项式复 杂性．全文共分五章，具体内容安排如下： 第一章概述内点算法的产生和研究现状以及相关的预备知识． 第二章是全文重点，主要讨论凸二次规划以及线性约束凸规划的宽邻域原始-对 偶势下降内点算法，并给出算法的多项式复杂性证明． 第三章讨论凸二次规划的一个 Mehrotra 型预估-校正算法及其改进算法，并给出 算法的多项式复杂性证明． 第四章给出初步的数值实验，验证本文算法的可行性以及有效性． 第五章对全文进行总结并对后续研究加以展望． 关键词：宽邻域内点算法 势函数 Scaled Lipschitz 条件 Mehrotra 型算法 多 项式复杂性 II Abstract Linear programming and nonlinear programming are two important branches of op- erations research, many practical problems can be characterized by them. Linear program- ming is a special case of convex programming, which is a special case of simple nonlinear programming. Therefore, it makes significant sense to study the algorithms for solving convex programming in both theory value and practical application. Convex quadratic pro- gramming is a simple kind of convex programming, researching the algorithms of it is not only for solving itself，but also for solving other nonlinear programs better, because the ordinary nonlinear functions can be well approximated by a quadratic function at their minimum points. The whole thesis majors in researching several interior-point algorithms in wide- neighborhood for convex quadratic programming, as well as analyzing their polynomial complexity. The whole thesis consists of five chapters and the concrete contents are ar- ranged as follows: The first chapter briefly introduces some historical backgrounds and the present re- search situations of interior poi