Newton's Method

From ProofWiki
Jump to navigation Jump to search

Proof Technique

Newton's method is a method of solving an equation expressed as a real function for which there may be no convenient closed form solution.

The derivative of the function has to be known in order to use Newton's method.

Let the equation to be solved be of the form:

$y = \map f x$

Let the value of $x$ be required for a given $y$.

Then an iterative improvement on an initial guess is of the form

$x_2 = x_1 - \dfrac {\map f {x_1} - y} {\map {f'} {x_1} }$

where $\map {f'} {x_1}$ is the derivative of $f$ with respect to $x$ evaluated at $x_1$.


Proof

The function $\map f x$ can be expanded using Taylor's Theorem:

$\map f {x_2} = \map f {x_1} + \map {f'} {x_1} \paren {x_2 - x_1} + \dfrac 1 2 \map {f''} {x_1} \paren {x_2 - x_1}^2 + \dotsb$

As $x_2$ gets closer to $x_1$, this series can be truncated to:

$\map f {x_2} = \map f {x_1} + \map {f'} {x_1} \paren {x_2 - x_1}$



Let $x_\infty$ be the exact solution where:

$\map f {x_\infty} = y$

Let $\epsilon \in \R_{>0}$ be the difference from the new estimate to the solution:

$x_\infty = x_2 + \epsilon$

Then the function expanded around the new estimate is:

$y = \map f {x_2} + \map {f'} {x_2} \epsilon$

Solving for $x_2$ produces:

$x_2 = x_1 - \dfrac {\map f {x_1} - y} {\map {f'} {x_1} } - \epsilon \dfrac {\map {f'} {x_2} } {\map {f'} {x_1} }$

For $\epsilon$ small enough the final term can be neglected:

$x_2 = x_1 - \dfrac {\map f {x_1} - y} {\map {f'} {x_1} }$



$\blacksquare$


Example

Newton's Method: $x^3 - 2 x - 5 = 0$

The real root of the cubic:

$x^3 - 2 x - 5 = 0$

can be found by using Newton's Method.

Its approximate value is:

$2 \cdotp 09455 \, 1$


Source of Name

This entry was named for Isaac Newton.


Sources