Newton's method

GUIDING QUESTION: How can I compute a solution to an equation?
Geometric motivation. Prime example of a fixed point method.

Motivation

Guiding philosophy: Dealing with arbitrary functions is hard. Solving a linear equation is easy! So let's boil down the root finding problem to solving some linear equations.

You'll soon see that reducing problems to
solving linear equations is a recurring theme in our course!

Motivation

We'll approximate a solution to a non-linear equation (hard) by iteratively solving a linear equation (easy).

Motivation

Suppose we are given $f$ and we would like to solve $f(x) = 0$. How could we use a linear equation to approximate a root $p$ of $f$?
Linearize $f$ using a Taylor approximation! When $x = x_n$ is close to $p$, the graph of $f$ looks a lot like its tangent line at $x_n$, so the point at which the tangent line crosses the $x$-axis is close to $p$.
This is the basis for our iterative scheme.
Using the Lagrange form of the remainder, Taylor's theorem implies that if $f$ is twice continuously differentiable on the interval $I$ between $x = x_n$ and $p$, then for each $x \in I$ there exists a number $\xi \in I$ such that $$f(x) = \color{var(--emphColor)}{f(x_n) + f'(x_n)(x - x_n)} + {f''(\xi) \over 2}(x - x_n)^2.$$ The tangent line to $f$ at $x = x_n$ omits the second-order remainder term proportional to $(x_n - x)^2$, so it is a first-order approximation.

$\pi$ according to Newton

Compute the tangent line at the current iterate and set your next iterate as the $x$-intercept of the tangent line.

Let's take a closer look

We may compute the tangent line and its zero analytically, thereby obtaining an explicit relationship between the current iterate and the next.

The tangent line to $f$ at $x = x_n$ is the unique line with slope $f'(x_n)$ containing $(x_n, f(x_n))$, so it is given by

$y = f(x_n) - f'(x_n)(x- x_n).$

Compare this expression to first terms in Taylor expansion of $f$ around $x = x_n$

$y = f(x_n) - f'(x_n)(x- x_n)$