文档详情

特征值与特征向量算法的研究.doc

发布:2017-11-24约1.22万字共18页下载文档
文本预览下载声明
特征值与特征向量算法的研究 摘 要: 目前在很多科学领域中进行研究时,问题常会转化成特征值与特征向量的求解。本文就求解特征值、特征向量的几个重要的算法作出了研究。如:幂法,反幂法,QR算法,Jacobi迭代法等。讨论了各算法的原理及各算法在MATLAB中的运行情况,从而比较出在面对不同性质的矩阵时每个算法都各有千秋。幂法计算简单,特别适用于高阶稀疏矩阵,但其收敛速度较慢,要想加快幂法的收敛速度可采用反幂法及位移技术。QR方法被人们称为数值数学,最值得注意的算法之一,它是目前求任意矩阵全部特征值和特征向量最有效的方法。Jacobi方法是古典方法,它收敛快,精度高,便于并行计算且算法稳定。但比较适用于求低阶的对称矩阵的全部特征值和特征向量。 关键词:特征值 特征向量 幂法 QR算法 雅可比算法 Abstract: At present While carrying on research in a lot of scientific fields,the questions often change into how to solve the eigenvalue and eigenvector. The degree paper do research in some important arithmetic on eigenvalue and eigenvector, such as power method, inverse power method, QR arithmetic and Jacobi arithmetic etc. In this paper, we discuss the theory of arithmetic, also including how to use them in the MATLAB. Then we can come to the conclusion that the power method is easy to run, it is fit to sparse matrix, but the speed is too slow! If you want to speed the rate of convergence, you can use inverse power method. QR is diffused as numerical mathematics, one of the noteworthiness arithmetic; it is the best arithmetic which can solve all eigenvalue and eigenvector of any matrix. Jacobin arithmetic is a classicality, the rate of convergence is fast, and the precision are high too. It is easy to parallel calculate, and the result is steady but it is fit to calculate all eigenvalue and eigenvector of symmetric matrix. Keywords:Eigenvalue eigenvector power method QR arithmetic Jacobin arithmetic 目录 摘要………………………………………………………………………(1) 1绪论 1.1问题提出与研究的目的和意义………………………………(3) 1.2国内外研究现状………………………………………………(3) 1.3论文结构与研究方法…………………………………………(3) 1.4论文使用的软件环境…………………………………………(4) 2 MATLAB语言及其在数值计算方面应用的简介…………………… (4) 2.1幂法……………………………………………………………(4) 2.2反幂法…………………………………………………………(6) 2.3移位反幂法……………………………………………………(8) 2.4 QR算法……………………………………………………… (10) 2.5雅可比(Jacobi)迭代法……………………………………(12) 3记单侧旋转法的对称矩阵特征值的求法……………………………(16) 4几种算法的比较………………………………………………………(16) 5 MATLAB计算仿真结果……………………………………………… (17) 在MATLAB中用幂法求其特征值与特征向量………………………(17) 6尚待深入研究
显示全部
相似文档