项目名称: 解约束优化问题的光滑化同伦方法研究
项目编号: No.11201240
项目类型: 青年科学基金项目
立项/批准年度: 2013
项目学科: 数理科学和化学
项目作者: 范晓娜
作者单位: 南京邮电大学
项目金额: 22万元
中文摘要: 同伦方法是一种重要的全局收敛性方法,其主要好处是能够在较弱的解存在性条件下得到大范围收敛性。本项目旨在运用光滑化同伦方法来解决一些特殊约束的数学规划问题。主要内容包括:(1)利用光滑化同伦方法,通过构造合适的同伦方程,来解决二阶锥的优化问题、半无限规划问题等,使得能够在较弱的解存在性条件下, 证明同伦路径的存在性和收敛性。这种方法与传统的组合同伦方法相比,由于不需要引进乘子变量,故将有更高的计算效率;(2)对于(1)中给出的同伦方法,结合同伦方程的特点,给出一个新的更有效的预估校正算法, 使之具有全局收敛性及多项式复杂性;(3)在其全局收敛性保证的条件下,利用预估校正方法,设计一个具有高阶局部收敛性的路径跟踪算法。通过以上内容的研究,本项目将为解决约束优化问题提供有效的新方法。
中文关键词: 约束优化;同伦方法;光滑化方法;全局收敛性;
英文摘要: Homotopy method is an important glabally convergent one. A distinctive advantage of the homotopy method is that the algorithm generated by it exhibits the global convergence under weaker conditions. The project aims to solve the special constrained mathematical programming problems by the smoothing homotopy method. The main contents include:(1)By constructing the suitable homotopy equation to solve the optimization problems with second-order cone programming,semi-infinite programming, and so forth via the smoothing homotopy method. Existence and convergence of the homotopy pathway are proven under some weaker solution conditions. The proposed method does not introduce in multiplier variables and hence it is more efficient than the combined homotopy method based on the KKT system. (2)A new predictor-corrector algorithm for tracing homotopy path will be proposed for (1), and its global convergence and polynomial complexity are established under some conditions.(3)To design a higher order lacally convergent algorithm with predictor step and corrector step for tracing the homotopy path. Throughout the above study, the project will provide a new and efficient method to solve the constrained optimization problems.
英文关键词: constrained optimization;homotopy method;smoothing method;global convergence;