2.4 error analysis for iterative methodszxu2/acms40390f12/lec-2.4.pdfΒ Β· newton’s method as fixed...

9
2.4 Error Analysis for Iterative Methods 1

Upload: others

Post on 03-Aug-2020

0 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: 2.4 Error Analysis for Iterative Methodszxu2/acms40390F12/Lec-2.4.pdfΒ Β· Newton’s Method as Fixed Point Problem Solve π‘₯= r by fixed point method. We write the problem as an equivalent

2.4 Error Analysis for Iterative Methods

1

Page 2: 2.4 Error Analysis for Iterative Methodszxu2/acms40390F12/Lec-2.4.pdfΒ Β· Newton’s Method as Fixed Point Problem Solve π‘₯= r by fixed point method. We write the problem as an equivalent

Definition β€’ Order of Convergence Suppose {𝑝𝑛}𝑛=0

∞ is a sequence that converges to 𝑝 with 𝑝𝑛 β‰  𝑝 for all 𝑛. If positive constants πœ† and 𝛼 exist with

limπ‘›β†’βˆž

|𝑝𝑛+1 βˆ’ 𝑝|

|𝑝𝑛 βˆ’ 𝑝|𝛼= πœ†

then {𝑝𝑛}𝑛=0∞ is said to converges to 𝒑 of order 𝜢 with asymptotic

error constant 𝝀. An iterative technique 𝑝𝑛 = 𝑔(π‘π‘›βˆ’1) is said to be of order 𝜢 if the sequence {𝑝𝑛}𝑛=0

∞ converges to the solution 𝑝 = 𝑔(𝑝) of order 𝜢. β€’ Special cases

1. If 𝛼 = 1 (and πœ† < 1), the sequence is linearly convergent 2. If 𝛼 = 2, the sequence is quadratically convergent 3. If 𝛼 < 1, the sequence is sub-linearly convergent (undesirable, very slow) 4. If 𝛼 = 1 and πœ† = 0 or 1 < 𝛼 < 2, the sequence is super-linearly convergent

β€’ Remark: High order (𝛼) ⟹ faster convergence (more desirable) πœ† is less important than the order (𝛼)

2

Page 3: 2.4 Error Analysis for Iterative Methodszxu2/acms40390F12/Lec-2.4.pdfΒ Β· Newton’s Method as Fixed Point Problem Solve π‘₯= r by fixed point method. We write the problem as an equivalent

Linear vs. Quadratic

3

Suppose we have two sequences converging to 0 with:

limπ‘›β†’βˆž

|𝑝𝑛+1|

|𝑝𝑛|= 0.9, lim

π‘›β†’βˆž

|π‘žπ‘›+1|

|π‘žπ‘›|2= 0.9

Roughly we have: 𝑝𝑛 β‰ˆ 0.9 π‘π‘›βˆ’1 β‰ˆ β‹― β‰ˆ 0.9𝑛 𝑝0 ,

π‘žπ‘› β‰ˆ 0.9|π‘žπ‘›βˆ’1|2 β‰ˆ β‹― β‰ˆ 0.92π‘›βˆ’1 π‘ž0 , Assume 𝑝0 = π‘ž0 = 1

Page 4: 2.4 Error Analysis for Iterative Methodszxu2/acms40390F12/Lec-2.4.pdfΒ Β· Newton’s Method as Fixed Point Problem Solve π‘₯= r by fixed point method. We write the problem as an equivalent

Fixed Point Convergence β€’ Theorem

Let 𝑔 ∈ 𝐢[π‘Ž, 𝑏] be such that 𝑔 π‘₯ ∈ [π‘Ž, 𝑏] for all π‘₯ ∈ π‘Ž, 𝑏 . Suppose 𝑔′ is continuous on (π‘Ž, 𝑏) and that 0 < π‘˜ < 1 exists with |𝑔′(π‘₯)| ≀ π‘˜ for all π‘₯ ∈ π‘Ž, 𝑏 .

If 𝑔′(𝑝) β‰  0, then for all number 𝑝0 in [π‘Ž, 𝑏], the sequence 𝑝𝑛 = 𝑔(π‘π‘›βˆ’1) converges only linearly to the unique fixed point 𝑝 in π‘Ž, 𝑏 .

β€’ Proof: 𝑝𝑛+1 βˆ’ 𝑝 = 𝑔 𝑝𝑛 βˆ’ 𝑔 𝑝 = 𝑔′ πœ‰π‘› 𝑝𝑛 βˆ’ 𝑝 , πœ‰π‘› ∈ (𝑝𝑛, 𝑝)

Since {𝑝𝑛}𝑛=0∞ converges to 𝑝, {πœ‰π‘›}𝑛=0

∞ converges to 𝑝.

Since 𝑔′ is continuous, limπ‘›β†’βˆž

𝑔′(πœ‰π‘›) = 𝑔′(𝑝)

limπ‘›β†’βˆž

|𝑝𝑛+1βˆ’π‘|

|π‘π‘›βˆ’π‘|= lim

π‘›β†’βˆžπ‘”β€² πœ‰π‘› = |𝑔′(𝑝)| ⟹ linear convergence

4

Page 5: 2.4 Error Analysis for Iterative Methodszxu2/acms40390F12/Lec-2.4.pdfΒ Β· Newton’s Method as Fixed Point Problem Solve π‘₯= r by fixed point method. We write the problem as an equivalent

Speed up Convergence of Fixed Point Iteration

β€’ If we look for faster convergence methods, we must have 𝑔′ 𝑝 = 0

β€’ Theorem Let 𝑝 be a solution of π‘₯ = 𝑔 π‘₯ . Suppose 𝑔′ 𝑝 = 0 and 𝑔′′ is continuous with 𝑔′′ π‘₯ < 𝑀 on an open interval 𝐼 containing 𝑝. Then there exists a 𝛿 > 0 such that for 𝑝0 ∈ 𝑝 βˆ’ 𝛿, 𝑝 + 𝛿 , the sequence defined by 𝑝𝑛+1 =𝑔 𝑝𝑛 , when 𝑛 β‰₯ 0, converges at least quadratically to 𝑝. For sufficiently large 𝑛

𝑝𝑛+1 βˆ’ 𝑝 <𝑀

2|𝑝𝑛 βˆ’ 𝑝|2

Remark:

Look for quadratically convergent fixed point methods which 𝑔 𝑝 = 𝑝 and 𝑔′ 𝑝 = 0.

5

Page 6: 2.4 Error Analysis for Iterative Methodszxu2/acms40390F12/Lec-2.4.pdfΒ Β· Newton’s Method as Fixed Point Problem Solve π‘₯= r by fixed point method. We write the problem as an equivalent

Newton’s Method as Fixed Point Problem

Solve 𝑓 π‘₯ = 0 by fixed point method. We write the problem as an equivalent fixed point problem:

𝑔 π‘₯ = π‘₯ βˆ’ 𝑓 π‘₯ solve: π‘₯ = 𝑔(π‘₯) 𝑔 π‘₯ = π‘₯ βˆ’ 𝛼𝑓 π‘₯ solve π‘₯ = 𝑔 π‘₯ , 𝛼 is a constant

𝑔 π‘₯ = π‘₯ βˆ’ πœ™ π‘₯ 𝑓 π‘₯ solve π‘₯ = 𝑔 π‘₯ , πœ™ π‘₯ is differentiable

Newton’s method is derived by the last form:

Find differentiable πœ™ π‘₯ with 𝑔′ 𝑝 = 0 when 𝑓 𝑝 = 0.

𝑔′ π‘₯ =𝑑

𝑑π‘₯π‘₯ βˆ’ πœ™ π‘₯ 𝑓 π‘₯ = 1 βˆ’ πœ™β€²π‘“ βˆ’ πœ™π‘“β€²

Use 𝑔′ 𝑝 = 0 when 𝑓 𝑝 = 0 𝑔′ 𝑝 = 1 βˆ’ πœ™β€² 𝑝 βˆ™ 0 βˆ’ πœ™ 𝑝 𝑓′ 𝑝 = 0

πœ™ 𝑝 = 1/𝑓′(𝑝)

This gives Newton’s method

𝑝𝑛+1 = 𝑔 𝑝𝑛 = 𝑝𝑛 βˆ’π‘“(𝑝𝑛)

𝑓′(𝑝𝑛)

6

Page 7: 2.4 Error Analysis for Iterative Methodszxu2/acms40390F12/Lec-2.4.pdfΒ Β· Newton’s Method as Fixed Point Problem Solve π‘₯= r by fixed point method. We write the problem as an equivalent

Multiple Roots

β€’ How to modify Newton’s method when 𝑓′ 𝑝 = 0. Here 𝑝 is the root of 𝑓 π‘₯ = 0.

β€’ Definition: Multiplicity of a Root A solution 𝑝 of 𝑓 π‘₯ = 0 is a zero of multiplicity π‘š of 𝑓 if for π‘₯ β‰  𝑝, we can write 𝑓 π‘₯ = π‘₯ βˆ’ 𝑝 π‘šπ‘ž π‘₯ , where limπ‘₯→𝑝 π‘ž(π‘₯) β‰  0.

β€’ Theorem 𝑓 ∈ 𝐢1[π‘Ž, 𝑏] has a simple zero at 𝑝 in (π‘Ž, 𝑏) if and only if 𝑓 𝑝 = 0, but 𝑓′(𝑝) β‰  0.

β€’ Theorem The function 𝑓 ∈ πΆπ‘š[π‘Ž, 𝑏] has a zero of multiplicity π‘š at point 𝑝 in (π‘Ž, 𝑏) if and only if 0 = 𝑓 𝑝 = 𝑓′ 𝑝 = 𝑓′′ 𝑝 = β‹― = 𝑓 π‘šβˆ’1 𝑝 , but 𝑓 π‘š (𝑝) β‰  0

7

Page 8: 2.4 Error Analysis for Iterative Methodszxu2/acms40390F12/Lec-2.4.pdfΒ Β· Newton’s Method as Fixed Point Problem Solve π‘₯= r by fixed point method. We write the problem as an equivalent

Newton’s Method for Zeroes of Higher Multiplicity (π‘š > 1)

Define the new function πœ‡ π‘₯ =𝑓(π‘₯)

𝑓′(π‘₯).

Write 𝑓 π‘₯ = π‘₯ βˆ’ 𝑝 π‘šπ‘ž(π‘₯), hence

πœ‡ π‘₯ =𝑓(π‘₯)

𝑓′(π‘₯)= π‘₯ βˆ’ 𝑝

π‘ž π‘₯

π‘šπ‘ž π‘₯ + π‘₯ βˆ’ 𝑝 π‘žβ€² π‘₯

Note that 𝑝 is a simple zero of πœ‡ π‘₯ .

β€’ Apply Newton’s method to πœ‡ π‘₯ to give:

π‘₯ = 𝑔 π‘₯ = π‘₯ βˆ’πœ‡ π‘₯

πœ‡β€² π‘₯

= π‘₯ βˆ’π‘“ π‘₯ 𝑓′ π‘₯

𝑓′ π‘₯ 2 βˆ’ 𝑓 π‘₯ 𝑓′′ π‘₯

β€’ Quadratic convergence: 𝑝𝑛+1 = 𝑝𝑛 βˆ’π‘“ 𝑝𝑛 𝑓′ 𝑝𝑛

𝑓′ 𝑝𝑛2βˆ’π‘“ 𝑝𝑛 𝑓′′ 𝑝𝑛

8

Page 9: 2.4 Error Analysis for Iterative Methodszxu2/acms40390F12/Lec-2.4.pdfΒ Β· Newton’s Method as Fixed Point Problem Solve π‘₯= r by fixed point method. We write the problem as an equivalent

Drawbacks:

β€’ Compute 𝑓′′(π‘₯) is expensive

β€’ Iteration formula is more complicated – more expensive to compute

β€’ Roundoff errors in denominator – both 𝑓′(π‘₯) and 𝑓(π‘₯) approach zero.

9