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.

and hence<br />

In conlcusion,<br />

∥<br />

∥I − J −1<br />

k J −J −1<br />

k+1∥ = k F(x(k+1) )∆x (k)<br />

∥ 2 ∥<br />

∥ ∥∆x (k) ∥∥ 2<br />

=<br />

(I − J −1 (∆x (k) ) T<br />

k J)∆x(k) ∥<br />

∥<br />

∥ ∥ ∥∆x (k) ∥∥ 2<br />

∥ ∥ 2 2 2 2<br />

∥<br />

≤ ∥I − J −1<br />

k J ∥∥2<br />

.<br />

(3.4.10) gives the ‖·‖ 2 -minimal relative correction of J k−1 , such that the secant condition (3.4.8)<br />

holds.<br />

△<br />

Fehlernorm<br />

10 0<br />

10 1<br />

10 −2<br />

10 0<br />

10 −4<br />

10 −6<br />

10 −1<br />

10 −8<br />

10 −2<br />

10 −10<br />

1 2 3 4 5 6 7 8 9 10 11<br />

Iterationsschritt<br />

Konvergenzmonitor<br />

Here:<br />

convergence monitor<br />

=<br />

quantity that displays difficulties in the<br />

convergence of an iteration<br />

∥<br />

∥J −1<br />

k−1 F(x(k) ) ∥<br />

µ :=<br />

∥<br />

∥∆x (k−1)∥ ∥ Heuristics: no convergence whenever µ > 1<br />

✸<br />

Remark 3.4.19. Option: damped Broyden method (as for the Newton method, section 3.4.4)<br />

△<br />

Ôº¿¾ ¿º<br />

Ôº¿¾ ¿º<br />

Example 3.4.18 (Broydens Quasi-Newton Method: Convergence).<br />

• In the non-linear system of the example 3.4.1, n = 2 take x (0) = (0.7.0.7) T and J 0 = DF(x (0) )<br />

The numerical example shows that the<br />

method is:<br />

slower than Newton method (3.4.1), but ✄<br />

Normen<br />

10 −2<br />

10 −4<br />

10 −6<br />

10 −8<br />

Implementation of (3.4.11): with Sherman-Morrison-Woodbury Update-Formula<br />

⎛<br />

⎞ ⎛<br />

⎞<br />

J −1<br />

k+1 = ⎜ J −1<br />

⎝I − k F(x(k+1) )(∆x (k) ) T ⎟<br />

∥<br />

∥<br />

∥∆x (k) ∥∥ 2<br />

⎠J −1 ⎜<br />

+ 2 ∆x(k) · J −1<br />

k = ⎝I + ∆x(k+1) (∆x (k) ) T ⎟<br />

∥<br />

k F(x(k+1) ∥<br />

)<br />

∥∆x (k) ∥∥ 2 ⎠J −1<br />

k<br />

2<br />

(3.4.12)<br />

that makes sense in the case that<br />

∥<br />

∥J −1<br />

k F(x(k+1) ∥<br />

) ∥ < ∥∆x (k)∥ ∥ ∥2<br />

2<br />

"simplified Quasi-Newton correction"<br />

10 0 Iterationsschritt<br />

better than simplified Newton method<br />

(see remark. 3.4.5)<br />

10 −10<br />

10 −12<br />

10 −14<br />

Broyden: ||F(x (k) )||<br />

Broyden: Fehlernorm<br />

Newton: ||F(x (k) )||<br />

Newton: Fehlernorm<br />

Newton (vereinfacht)<br />

0 1 2 3 4 5 6 7 8 9 10 11<br />

Ôº¿¾ ¿º<br />

Ôº¿¾ ¿º

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

Saved successfully!

Ooh no, something went wrong!