Math 4329: Numerical Analysis Chapter 03: Newton s Method. Natasha S. Sharma, PhD

Transcription

1

2 Mathematical question we are interested in numerically answering How to find the x-intercepts of a function f (x)? These x-intercepts are called the roots of the equation f (x) = 0. Notation: denote the exact root by α. That means, f (α) = 0.

3 Basic Idea Behind Given x 0, x 1 is the x-intercept of the tangent line at (x 0, f (x 0 )).

4 Tangent Line at (x 0, f (x 0 )): y(x) = f (x 0 ) + f (x 0 )(x x 0 ). We obtain the next iterate x 1 as the x-intercept of the tangent line that is f (x 0 ) + f (x 0 )(x 1 x 0 ) = 0. This simplifies to x 1 = x 0 f (x 0) f (x 0 ). Generalizing, we can generate a sequence {x n } n 1 where x n+1 = x n f (x n) f, n = 0, 1, 2, (x n )

5 Example Find the largest root of f (x) = x 6 x 1 = 0 accurate within ε = 1e 8 using.

6 Solution Note α Solution: The sequence of iterates {x n } n 1 is generated according to the formula: for all n = 0, 1, 2, ( x 6 x n+1 = x n n x n 1 ) 6xn 5, 1 ( 6x 5 = x n 1 ) ( x 6 n 6xn 5 n x n 1 ) 1 6xn 5 1 ( ) 6xn 6 x n xn 6 x n 1 = 6xn 5 1 = 5x 6 n + 1 6x 5 n 1.

7 Performance of the n x n f (x n ) x n x n 1 α x n e e+1-2e e e e e e-2-4.2e e e e e e e e e e e-9.. α Remarks 1 May converge slowly at first. However, as the iterates come closer to the root, the speed of convergence increases.

8 Another Example Using solve the following equation f (x) x 3 3x 2 + 3x 1 = 0 with an accuracy of ε = Simplified form of : with initial guess x 0 = 0.5. x n+1 = 2x 3 n x 2 n + 1 3(x n 1) 2,

9 Application I: Root finding in any dimension Example: Finding the intersection of a hyperbola and a circle. Intersection of a circle and a hyperbola 6 4 y coordinate x coordinate

10 Application II: Division Operation Replace the division operation in early computers. These early computers only allowed addition, subtraction and multiplication. Compute 1 b using and the operations +,,. Solution: Find x such that x = 1 b. Equivalently, find x satisfying f (x) := b x 1 = 0 : Start with initial guess x 0, compute x 1 using x 1 = x 0 (2 bx 0 ).

11 Error Assume that f (x) has atleast continuous derivatives of order 2 for all x in some interval containing α and f (α) 0. α x n+1 = (α x n ) 2[ f (c n ) ] 2f. (x n ) Error in x n+1 is nearly proportional to the square of the error in x n. The term f (c n) 2f (x n) is the amplification factor. However, it depends on n. We need to make this factor independent of n. This can be achieved in the following manner: f (c n ) 2f (x n ) f (α) 2f (α) = M. f (x) M = max x [a,b] 2f (x).

12 Error Initial guess is crucial here and determine the number of iterations needed to achieve the desired accuracy! For our worked out example, f (c n ) 2f (x n ) f (α) 2f (α) α x n (α x n ) 2

13 Determining x 0 without using Bisection α x n+1 = (α x n ) 2[ f (c n ) ] 2f (x n [ ) (α x n ) 2 f (α) ] 2f (α) }{{} M Multiplying both sides with M M(α x n+1 ) M 2 (α x n ) 2 M(α x 2 ) M 2 (α x 1 ) 2 M 2( M 2 (α x 0 ) 4) = ( ) 2 2 M(α x 0 ).

14 M(α x 0 ) < 1 = α x 0 < 1 M By picking x 0 1 < 1/b x 0 1/b < 1 1 < 1 bx 0 < < bx 0 < 2

15 Order of Convergence A sequence {x n } n 0 converges to α with order p 1 if α x n+1 c α x n p, n 0 for some c 0 p = 1 and c < 1 linear convergence (Bisection ), p = 2 quadratic convergence ( ), p = 3 cubic convergence (some fixed point iterative methods).