Paper 4, Section II, D

Numerical Analysis | Part II, 2014

Let AA be a real symmetric n×nn \times n matrix with nn distinct real eigenvalues λ1<λ2<\lambda_{1}<\lambda_{2}< <λn\cdots<\lambda_{n} and a corresponding orthogonal basis of normalized real eigenvectors {wi}i=1n\left\{\mathbf{w}_{i}\right\}_{i=1}^{n}.

(i) Let sRs \in \mathbb{R} satisfy s<λ1s<\lambda_{1}. Given a unit vector x(0)Rn\mathbf{x}^{(0)} \in \mathbb{R}^{n}, the iteration scheme

(AsI)y=x(k)x(k+1)=y/y\begin{gathered} (A-s I) \mathbf{y}=\mathbf{x}^{(k)} \\ \mathbf{x}^{(k+1)}=\mathbf{y} /\|\mathbf{y}\| \end{gathered}

generates a sequence of vectors x(k+1)\mathbf{x}^{(k+1)} for k=0,1,2,k=0,1,2, \ldots. Assuming that x(0)=ciwi\mathbf{x}^{(0)}=\sum c_{i} \mathbf{w}_{i} with c10c_{1} \neq 0, prove that x(k)\mathbf{x}^{(k)} tends to ±w1\pm \mathbf{w}_{1} as kk \rightarrow \infty. What happens to x(k)\mathbf{x}^{(k)} if s>λ1s>\lambda_{1} ? [Consider all cases.]

(ii) Describe how to implement an inverse-iteration algorithm to compute the eigenvalues and eigenvectors of AA, given some initial estimates for the eigenvalues.

(iii) Let n=2n=2. For iterates x(k)\mathbf{x}^{(k)} of an inverse-iteration algorithm with a fixed value of sλ1,λ2s \neq \lambda_{1}, \lambda_{2}, show that if

x(k)=(w1+ϵkw2)/(1+ϵk2)1/2\mathbf{x}^{(k)}=\left(\mathbf{w}_{1}+\epsilon_{k} \mathbf{w}_{2}\right) /\left(1+\epsilon_{k}^{2}\right)^{1 / 2}

where ϵk\left|\epsilon_{k}\right| is small, then ϵk+1\left|\epsilon_{k+1}\right| is of the same order of magnitude as ϵk\left|\epsilon_{k}\right|.

(iv) Let n=2n=2 still. Consider the iteration scheme

sk=(x(k),Ax(k)),(AskI)y=x(k),x(k+1)=y/ys_{k}=\left(\mathbf{x}^{(k)}, A \mathbf{x}^{(k)}\right), \quad\left(A-s_{k} I\right) \mathbf{y}=\mathbf{x}^{(k)}, \quad \mathbf{x}^{(k+1)}=\mathbf{y} /\|\mathbf{y}\|

for k=0,1,2,k=0,1,2, \ldots, where (,(,, denotes the inner product. Show that with this scheme ϵk+1=ϵk3.\left|\epsilon_{k+1}\right|=\left|\epsilon_{k}\right|^{3} .

Typos? Please submit corrections to this page on GitHub.