The [[Fixed Point|Fixed Point]] Method is an iterative method to find the [[Fixed Point|fixed point]] of a function $f$.
$\huge f: p \mapsto p $
The fixed point method is to simply keep applying the function.
$\huge x_{n+1} = f(x_{n}) $
>[!info] Restriction to guarentee this method works
>$\huge f\in \mathcal C^{1}[a,b] $
>$\huge
>\forall x\in [a,b]: |f'(x)| <1
>$
### Asymtopic Behavior
Solve for $p$ such that $f(p)=p$.
$\forall x, \let \hat{x} = x - p$.
$\huge \begin{align}
x_{n+1} &= f(x_{n})
\end{align}$
[[Taylor Series]] at $x=p$
$\large \begin{align}
f(x) &= f(p) +f'(p)(x-p)+ \frac{1}{2} f''(p)(x-p)^{2} + \cdots \\
&= p + f'(p) \hat{x} + \frac{1}{2} f''(p) \hat{x}^{2} + \cdots
\end{align}$
$\huge \begin{align}
x_{n+1} &= p + f'(p) \hat{x}_{n} + \frac{1}{2} + \mathcal O(\hat{x}_{n}^{2}) \\
\end{align}$
The [[Numerical Analysis Error Bounds|Error Bound]] will be:
$\huge
\epsilon_{n} = \left| \hat{x}_{n} \right|
$
$\huge \begin{align}
\hat{x}_{n+1} &= x_{n+1} - p \\
&= f'(p) \hat{x}_{n} + \mathcal O( \hat{x}_{n}^{2})
\end{align}$
As $n\to \infty$, $\hat{x}_{n}\to 0$, so we discard the negligent $\mathcal O$:
$\huge
\hat{x}_{n+1} \approx f'(p) \hat{x}_{n}
$
Because $f'(p)$ is constant, this means that the error of the fixed point method has an [[Order of Convergence]] $\alpha=1$.
### Approximation using [[Aitken's Delta Squared]]
Let $p$ be such that $f(p) =p$.
The taylor series of $f$ around $x=p$ is:
$\large
f(x) = p + f'(p)(x-p)+ \frac{1}{2}f''(x)(x-p)^{2}
+ \mathcal O((x-p)^{3})
$
$\huge \let \hat{x} = x-p $
$\large
\begin{align}
f(x) &= p + f'(p)\hat{x}+ \frac{1}{2}f''(p) \hat{x}^{2} + \mathcal O(\hat{x}^{3})\\
\end{align}
$
$\huge \begin{align}
f(x) &= p + f'(p)\hat{xi_{i}}+ \frac{1}{2}f''(p) \hat{x_{i}}^{2} + \mathcal O(\hat{x_{i}}^{3})\\
\hat{x}_{i+1} &= x_{i+1}-p \\
&= f'_{p} \hat{x}_{i} + \frac{1}{2}f_{p}''\hat{x}_{i}^{2} + \mathcal O(\hat{x}_{i}^{3}) \\
\Delta \hat{x}_{i} &= x_{i+1} - \hat{x}_{i}\\
&= (f_{p}'-1) \hat{x}_{i} + \frac{1}{2}f_{p}'' \hat{x}_{i}^{2} + \mathcal O( \hat{x}_{i}^{3}) \\
\Delta^{2} \hat{x}_{i} &=
(f_{p}' - 1)^{2} \hat{x}_{i} +
\frac{1}{2}\pa{ f_{p}'^{2} + {f'}_{p}^{1} - 2 } \hat{x}_{i}^{2} + \mathcal O (\hat{x}^{3}) \\
&=
\hat{x}_{i}
- 2. \frac{f_{p}'}{f_{p}'-1} f_{p}'' \hat{x}_{i}^{2} + \mathcal O(\hat{x}_{i}^{3})
\end{align}
$
$\huge
x_{i} - \frac{\Delta x_{i}^{2}}{\Delta^{2} x_{i}} =
p +
2 \frac{f_{p}'}{f_{p}'-1} f_{p}'' \hat{x}_{i}^{2} + \mathcal O(\hat{x}_{i}^{3})
$