Вы находитесь на странице: 1из 20

MATH 175: Numerical Analysis II

Mathematics Division
Institute of Mathematical Sciences and Physics
University of the Philippines Los Baños

1 / 20
Fixed Point

Definition
A fixed point of a function g(x) is a real number P such that P = g(P )

How is a fixed-point problem related to a root-finding problem?

Given a root -finding problem f (p) = 0, we can define a function g with


fixed point at p.

2 / 20
Fixed Point

Theorem
Assume that g ∈ C[a, b].
a. If g(x) ∈ [a, b] ∀x ∈ [a, b], then g has a fixed point in [a, b].
b. Furthermore, suppose that g 0 (x) exists on (a, b) and ∃K < 1 with
|g 0 (x)| ≤ K ∀x ∈ (a, b), then the fixed point in [a, b] is unique.

Proof done in class

3 / 20
Example

x2 − 1
Let g(x) = on [−1, 1].
3
Extreme Value Theorem implies that the absolute minimum value of g
1
occurs at x = 0 and g(0) = − . Similarly, the absolute maximum value of
3
g occurs at x = ±1 and has a value 0. Moreover, g is continuous and

0
2x 2
|g (x)| = ≤ , for all x ∈ [−1, 1]
3 3

Thus, g has a unique fixed point in [−1, 1]


The unique fixed point
√ in [−1, 1] can be determined algebraically which is
3 − 13
equivalent to
2

4 / 20
Example

Argue that h(x) = cos x has a unique fixed point on [−1, 1].

5 / 20
Fixed Point Iteration

Definition
The iteration pn+1 = g(pn ) for n = 0, 1, 2, ... is called fixed point
iteration.

6 / 20
Fixed Point Iteration

Theorem
Suppose that g is a continuous function and {pn }∞
n=0 is a sequence
generated by fixed-point iteration. If lim pn = P then P is a fixed point
n→∞
of g(x).

Proof:

lim pn = P =⇒ lim pn+1 = P


n→∞ n→∞

thus,
 
g(P ) = g lim pn = lim g(pn ) = lim pn+1 = P
n→∞ n→∞ n→∞

∴ P is a fixed point of g(x).


7 / 20
Fixed Point Iteration

How to we visualize the fixed point iteration process?

8 / 20
Example

Consider the equation x3 + 4x2 − 10 = 0. The equation has a unique root


in [1, 2].
Below are some iteration functions whose fixed points are roots of the
given equation.
a. x = g1 (x) = x − x3 − 4x2 + 10
 1/2
10
b. x = g2 (x) = − 4x
x
1 1/2
c. x = g3 (x) = 10 − x3
2
10 1/2
 
d. x = g4 (x) =
4+x
x3 + 4x2 − 10
e. x = g5 (x) = x −
3x2 + 8x

9 / 20
Example

10 / 20
Fixed-point Theorem
Suppose that
(i) g ∈ C[a, b] and g 0 ∈ C[a, b]
(ii) K ≥ 0
(iii) p0 ∈ (a, b)
(iv) g(x) ∈ [a, b] ∀x ∈ [a, b]

a. If |g 0 (x)| ≤ K < 1 ∀x ∈ [a, b], then the iteration pn = g(pn−1 ) will


converge to the unique fixed point P ∈ [a, b].
b. If |g 0 (x)| > 1 ∀x ∈ [a, b], then the iteration pn = g(pn−1 ) will not
converge to P .

Proof was shown in class.

11 / 20
Example
a. x = g1 (x) = x − x3 − 4x2 + 10

When g1 (x) = x − x3 − 4x2 + 10, g10 (x) = 1 − 3x2 − 8x. There is no


interval [a, b] containing p for which |g10 (x)| < 1. Hence, there is no
reason to expect convergence.
12 / 20
Example

 1/2
10
b. x = g2 (x) = − 4x
x

13 / 20
Example

 1/2
10
With g2 (x) = − 4x , we can see that g2 does not map [1, 2] into
x
[1, 2], and the sequence {pn }∞n=0 is not defined with p0 = 1.5. Moreover,
there is no interval containing p such that

|g20 (x)| < 1, since |g20 (p)| ≈ 3.4

14 / 20
Example

1 1/2
c. x = g3 (x) = 10 − x3
2

15 / 20
Example
1
1/2
For the function g3 (x) = 2 10 − x3 ,

3
g30 (x) = − x2 (10 − x3 )−1/2 < 0 on [1, 2]
4
so g3 is strictly decreasing on [1, 2]. However, |g30 (2)| ≈ 2.12, so the
condition |g30 (x)| ≤ k fails on [1, 2]. A closer examination of the sequence
{pn }∞n=0 starting with p0 = 1.5 shows that it suffices to consider the
interval [1, 1.5] instead of [1, 2]. On this interval it is still true that
g30 (x) < 0 and g3 is strictly decreasing, but, additionally,

1 < 1.28 ≈ g3 (1.5) ≤ g3 (1) = 1.5

for all x ∈ [1, 1.5]. This shows that g3 maps the interval [1, 1.5] into itself.
Since it is also true that |g30 (x)| ≤ |g30 (1.5)| ≈ 0.66 on this interval, thus g3
converge to p.
16 / 20
Example

 1/2
10
d. x = g4 (x) =
4+x

17 / 20
Example

10 1/2
 
For g4 (x) = we have
4+x

0
−5 ≤ √ 5

|g4 (x)| =
√ < 0.15, for all x ∈ [1, 2]
3/2
10(4 + x) 1053/2

The bound on the magnitude of g40 (x) is much smaller than the bound on
the magnitude of g30 (x), which explains the more rapid convergence using
g4 .

18 / 20
Example

x3 + 4x2 − 10
e. x = g5 (x) = x −
3x2 + 8x

19 / 20
Error Bound

Corollary
Assume that g satisfies the hypothesis given in the previous theorem.
Bounds for the error involved when using pn to approximate P are given by

|P − pn | ≤ K n max{p0 − a, b − p0 } for all n ≥ 1

and
K n |p1 − p0 |
|P − pn | ≤ for all n ≥ 1
1−K

Proof was shown in class.

20 / 20

Вам также может понравиться