21.06.2014 Views

Numerical Methods Contents - SAM

Numerical Methods Contents - SAM

Numerical Methods Contents - SAM

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

(x (k) ) k∈N0 Cauchy sequence ➤ convergent x (k) k→∞<br />

−−−−→ x ∗ . .<br />

Continuity of Φ ➤ Φ(x ∗ ) = x ∗ . Uniqueness of fixed point is evident. ✷<br />

△<br />

Φ(x)<br />

Φ(x)<br />

A simple criterion for a differentiable Φ to be contractive:<br />

✬<br />

✩<br />

x<br />

x<br />

Lemma 3.2.4 (Sufficient condition for local linear convergence of fixed point iteration).<br />

If Φ : U ⊂ R n ↦→ R n , Φ(x ∗ ) = x ∗ ,Φ differentiable in x ∗ , and ‖DΦ(x ∗ )‖ < 1, then the fixed<br />

point iteration (3.2.1) converges locally and at least linearly.<br />

0 ≤ Φ ′ (x ∗ ) < 1 ➣ convergence<br />

1 < Φ ′ (x ∗ ) ➣ divergence<br />

✸<br />

✫<br />

matrix norm, Def. 2.5.2 !<br />

✪<br />

Proof. (of Lemma 3.2.4)<br />

By definition of derivative<br />

✎ notation: DΦ(x) ˆ= Jacobian (ger.: Jacobi-Matrix) of Φ at x ∈ D<br />

→ [40, Sect. 7.6]<br />

Example 3.2.3 (Fixed point iteration in 1D).<br />

1D setting (n = 1):<br />

Ôº¾ ¿º¾<br />

Φ : R ↦→ R continuously differentiable, Φ(x ∗ ) = x ∗<br />

fixed point iteration: x (k+1) = Φ(x (k) )<br />

with ψ : R + 0 ↦→ R+ 0<br />

Choose δ > 0 such that<br />

‖Φ(y) − Φ(x ∗ ) − DΦ(x ∗ )(y − x ∗ )‖ ≤ ψ(‖y − x ∗ ‖) ‖y − x ∗ ‖ ,<br />

satisfying lim ψ(t) = 0.<br />

t→0<br />

L := ψ(t) + ‖DΦ(x ∗ )‖ ≤ 1 2 (1 + ‖DΦ(x∗ )‖) < 1 ∀0 ≤ t < δ .<br />

Ôº¾ ¿º¾<br />

“Visualization” of the statement of Lemma 3.2.4: The iteration converges locally, if Φ is flat in a<br />

neighborhood of x ∗ , it will diverge, if Φ is steep there.<br />

Φ(x)<br />

Φ(x)<br />

By inverse triangle inequality we obtain for fixed point iteration<br />

‖Φ(x) − x ∗ ‖ − ‖DΦ(x ∗ )(x − x ∗ )‖ ≤ ψ(‖x − x ∗ ‖) ‖x − x ∗ ‖<br />

∥<br />

∥x (k+1) − x ∗∥ ∥ ≤ (ψ(t) + ‖DΦ(x ∗ ∥<br />

)‖) ∥x (k) − x ∗∥ ∥<br />

∥ ∥∥x ≤ L (k) − x ∗∥ ∥ ,<br />

∥<br />

if ∥x (k) − x ∗∥ ∥ < δ.<br />

✷<br />

Contractivity also guarantees the uniqueness of a fixed point, see the next Lemma.<br />

x<br />

x<br />

Recalling the Banach fixed point theorem Thm. 3.2.3 we see that under some additional (usually mild)<br />

assumptions, it also ensures the existence of a fixed point.<br />

−1 < Φ ′ (x ∗ ) ≤ 0 ➣ convergence<br />

Φ ′ (x ∗ ) < −1 ➣ divergence<br />

✬<br />

✩<br />

Ôº¾ ¿º¾<br />

Lemma 3.2.5 (Sufficient condition for local linear convergence of fixed point iteration (II)).<br />

Let U be convex and Φ : U ⊂ R n ↦→ R n be continuously differentiable with L :=<br />

sup ‖DΦ(x)‖ < 1. If Φ(x ∗ ) = x ∗ for some interior point x ∗ ∈ U, then the fixed point iteration<br />

x∈U<br />

x (k+1) = Φ(x (k) ) converges to x ∗ locally at least linearly.<br />

✫<br />

Ôº¾ ¿º¾<br />

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

Saved successfully!

Ooh no, something went wrong!