项目名称: 基于二次规划的大规模非线性半定规划问题的理论、算法研究及软件设计
项目编号: No.11201382
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: 王承竞
作者单位: 西南交通大学
项目金额: 22万元
中文摘要: 本项目拟研究几类重要的基于二次规划的大规模非线性半定规划问题:(1).凸二次目标函数加上entropy项的半定规划问题;(2).凸二次目标函数加上log-determinant项的半定规划问题;(3).带秩约束的二次半定规划问题。这些问题在信息、统计、金融、图像处理等领域有着广泛的应用。几类问题之间虽然形式各有差别,但从问题类型和解决途径上又有不少相似之处。为了求解这些问题,我们将充分挖掘问题本身的特点,拟采用不精确的一阶和二阶方法混合使用的办法,设计复杂度为O(1/k^2)的高效数值算法,编写应用软件。对这几类特殊大规模非线性半定规划问题的研究,必能为更一般的非线性半定规划问题的解决提供新的工具、方法和思想。
中文关键词: 非线性半定规划;二次规划;邻近增广拉格朗日方法;增广拉格朗日方法;Faài Bruno公式
英文摘要: Our project aims at the research on several important classes of large scale nonlinear semidefinite programming problems (SDPs) based on quadratic programmings: (1).SDPs whose objective is a covex quadratic function with an entropy term; (2).SDPs whose objective is a convex quadratic function with a log-determinant term; (3).quadratic SDPs with rank constraints. These problems can find wide applications in the areas of information, statistics, finance, image processing, and so on. They are also similar to each other in the problem types and the ways in which to solve them though they have different appearances. In order to solve these problems, we will fully take advantage of the characteristics of the problems themselves, design efficient numerical algorithms whose complexity is O(1/k^2) with the idea of the combination of the first-order and second-order methods,and make application software. The research on these special large scale nonlinear SDPs will definitely provide new tools, methods and ideas for more general nonlinear SDPs.
英文关键词: nonlinear semidefinite programming;quadratic programming;proximal augmented Lagrangian method;Faài Bruno's formula;