Paper 1, Section II, E

Numerical Analysis | Part II, 2020

Let ARn×nA \in \mathbb{R}^{n \times n} be a real symmetric matrix with distinct eigenvalues λ1<λ2<<\lambda_{1}<\lambda_{2}<\cdots< λn\lambda_{n} and a corresponding orthonormal basis of real eigenvectors {wi}i=1n\left\{\boldsymbol{w}_{i}\right\}_{i=1}^{n}. Given a unit norm vector x(0)Rn\boldsymbol{x}^{(0)} \in \mathbb{R}^{n}, and a set of parameters skRs_{k} \in \mathbb{R}, consider the inverse iteration algorithm

(AskI)y=x(k),x(k+1)=y/y,k0.\left(A-s_{k} I\right) \boldsymbol{y}=\boldsymbol{x}^{(k)}, \quad \boldsymbol{x}^{(k+1)}=\boldsymbol{y} /\|\boldsymbol{y}\|, \quad k \geqslant 0 .

(a) Let sk=s=s_{k}=s= const for all kk. Assuming that x(0)=i=1nciwi\boldsymbol{x}^{(0)}=\sum_{i=1}^{n} c_{i} \boldsymbol{w}_{i} with all ci0c_{i} \neq 0, prove that

s<λ1x(k)w1 or x(k)w1(k).s<\lambda_{1} \Rightarrow \boldsymbol{x}^{(k)} \rightarrow \boldsymbol{w}_{1} \quad \text { or } \quad \boldsymbol{x}^{(k)} \rightarrow-\boldsymbol{w}_{1} \quad(k \rightarrow \infty) .

Explain briefly what happens to x(k)\boldsymbol{x}^{(k)} when λm<s<λm+1\lambda_{m}<s<\lambda_{m+1} for some m{1,2,,n1}m \in\{1,2, \ldots, n-1\}, and when λn<s\lambda_{n}<s.

(b) Let sk=(Ax(k),x(k))s_{k}=\left(A \boldsymbol{x}^{(k)}, \boldsymbol{x}^{(k)}\right) for k0k \geqslant 0. Assuming that, for some kk, some aiRa_{i} \in \mathbb{R} and for a small ϵ\epsilon,

x(k)=c1(w1+ϵi2aiwi)\boldsymbol{x}^{(k)}=c^{-1}\left(\boldsymbol{w}_{1}+\epsilon \sum_{i \geqslant 2} a_{i} \boldsymbol{w}_{i}\right)

where cc is the appropriate normalising constant. Show that sk=λ1Kϵ2+O(ϵ4)s_{k}=\lambda_{1}-K \epsilon^{2}+\mathcal{O}\left(\epsilon^{4}\right) and determine the value of KK. Hence show that

x(k+1)=c11(w1+ϵ3i2biwi+O(ϵ5))\boldsymbol{x}^{(k+1)}=c_{1}^{-1}\left(\boldsymbol{w}_{1}+\epsilon^{3} \sum_{i \geqslant 2} b_{i} \boldsymbol{w}_{i}+\mathcal{O}\left(\epsilon^{5}\right)\right)

where c1c_{1} is the appropriate normalising constant, and find expressions for bib_{i}.

Typos? Please submit corrections to this page on GitHub.