图Graph-北京大学.PDF

求解特征值的数值方法简介 卢朓 tlu@math.pku.edu.cn School of Mathematical Sciences, Peking University Eigenvalue and Eigenvector A∈Rn×nis a squrae matrix n If there is a vector x ∈R , such that Ax =x for some scalar , then  is called the eigenvalue ofA with corresponding (right) eigenvector x 特征多项式 A−Ix =0 is an eigenvalue of A means the above has linear equation nonzero solution. 特征多项式 det A−I=0 求特征值转化成求特征多项式的根 Problem Transformations • Shift : for scalar  , eigenvalues of A−I are eigenvalues • of A shifted by   −i ­1 • Inversion : for nonsingular A, eigenvalues of A  are reciprocals of eigenvalues of A 1/i k • Powers : for integer k  0, eigenvalues of A  are kth powers of eigenvalues of A ki • Polynomial : for polynomial p(t), eigenvalues of p(A) are • values of p evaluated at eigenvalues of A, p i Similarity Transformations B is similar to A if there is nonsingular T such that −1 B=T A T Then   −1     By =y ⇒T AT y =y ⇒AT y=T y  so A and B have same eigenvalues, and if y is  eigenvector of B, then x=T y is eigenvector of A Similarity transformations preserve eigenvalues, and eigenvectors are easily recovered Michael T. Heath Parallel Numeri Forms attainable by similarity transformation A T