03.03.2014 Views

Numerical Methods Course Notes Version 0.1 (UCSD Math 174, Fall ...

Numerical Methods Course Notes Version 0.1 (UCSD Math 174, Fall ...

Numerical Methods Course Notes Version 0.1 (UCSD Math 174, Fall ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

3.4. ITERATIVE SOLUTIONS 45<br />

is the matrix<br />

⎡<br />

⎢<br />

⎣<br />

1 0 0 · · · 0<br />

−a 2 1 0 · · · 0<br />

−a 3 0 1 · · · 0<br />

. . .<br />

. .. .<br />

−a n 0 0 · · · 1<br />

(Hint: Multiply them together.)<br />

(3.7) Under the strategy of scaled partial pivoting, which row of the following matrix will be the<br />

first pivot row?<br />

⎡<br />

⎤<br />

10 17 −10 <strong>0.1</strong> 0.9<br />

−3 3 −3 0.3 −4<br />

⎢ 0.3 <strong>0.1</strong> 0.01 −1 0.5<br />

⎥<br />

⎣ 2 3 4 −3 5 ⎦<br />

10 100 1 <strong>0.1</strong> 0<br />

(3.8) Let A be a symmetric positive definite n × n matrix with n distinct eigenvalues. Letting<br />

y (0) = b/ ‖b‖ 2<br />

, consider the iteration<br />

y (k+1) =<br />

Ay(k) ∥ Ay (k) ∥ ∥<br />

2<br />

.<br />

(a) What is ∥ ∥ y<br />

(k) ∥ ∥<br />

2<br />

?<br />

(b) Show that y (k) = A k b/ ∥ ∥ A k b ∥ ∥<br />

2<br />

.<br />

(c) Show that as k → ∞, y (k) converges to the (normalized) eigenvector associated with the<br />

largest eigenvalue of A.<br />

(3.9) Consider the equation<br />

⎡<br />

⎣<br />

1 3 5<br />

−2 2 4<br />

4 −3 −4<br />

⎤<br />

⎦ x = ⎣<br />

Letting x (0) = [1 1 0] ⊤ , find the iterate x (1) by one step of Richardson’s Method. And by<br />

one step of Jacobi Iteration. And by Gauss Seidel.<br />

(3.10) Let A be a symmetric n × n matrix with eigenvalues in the interval [α, β], with 0 < α ≤ β,<br />

and α + β ≠ 0. Consider Richardson’s Iteration<br />

⎡<br />

⎤<br />

⎥<br />

⎦<br />

−5<br />

−6<br />

10<br />

x (k+1) = (I − ωA) x (k) + ωb.<br />

Recall that e (k+1) = (I − ωA) e (k) .<br />

(a) Show that the eigenvalues of I − ωA are in the interval [1 − ωβ, 1 − ωα].<br />

(b) Prove that<br />

max {|λ| : 1 − ωβ ≤ λ ≤ 1 − ωα}<br />

is minimized when we choose ω such that 1 − ωβ = − (1 − ωα) . (Hint: It may help to<br />

look at the graph of something versus ω.)<br />

(c) Show that this relationship is satisfied by ω = 2/ (α + β).<br />

(d) For this choice of ω show that the spectral radius of I − ωA is<br />

|α − β|<br />

|α + β| .<br />

⎤<br />

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!