Research on Algorithms for least squares and linear constrained optimization problems最小二乘和线性约束优化问题的算法研究.pdfVIP

最小二乘和线性约束优化问题的算法研究 摘要 本文主要是研究最小二乘和线性约束优化问题的一些数值算法,全文总共 四章内 容，安排如下： 第一章，主要介绍了最小二乘和线性约束优化问题的基本概念及相关基础知识，并且 给出了本文主要结果。 第二章，主要是给出了一类箱约束线性最小二乘问题的牛顿算法，在适当条件下，证 明这种算法的有限步终止性。 第三章，主要建立了带线性约束的凸规划问题的一个新型增广Lagrange 算法，并通过 引入新参数对该算法进行加速。理论上证明出新型算法的收敛速率和计算复杂性均为 O(1k )，而加速后的算法在每一迭代步的计算复杂性为O( 1k 2 )。 第四章，主要总结本文的主要结果，指出今后研究方向。 关键词 箱约束最小二乘问题；有限步终止；牛顿算法；凸规划；增广Lagrange 方法；加速技术。 Research About the Least-square and The Linearly Constrained Convex Optimization Problem ABSTRACT This paper mainly studies numerical algorithm for the least-square and the linearly constrained convex optimization problem. The structure of this paper is organized as follows: In chapter 1, the basic concepts and related basic knowledge of the least-square and the linearly constrained convex optimization problem are introduced. Then, we summarize the main results of this paper. In chapter 2, we mainly research the finite step termination of the Newton method for the boxed constrained linear least-square problem under proper conditions. In chapter 3, we first establish a new augmented Lagrange method for a linearly constrained convex optimization, and then introduce an acceleration technique. It is shown that the 1 computing complexity and convergence rate of the augmented Lagrange method are all O( k ), and the computing complexity of the accelerated new augmented Lagrange method is O( 1k 2 ). In chapter 4, the main results of the paper are summarized, and points out future research directions. KEYWORDS Boxed constrained least-square problem; Finite-step termination; Newton method; Convex programming; Augmented