Recap
Convergence Order
For some iteration algorithm x(t+1)=g(x(t)) with final solution denoted x∗, error ε(t):=x(t)−x∗, its convergence order α and converngence rate c:
t→∞lim∣ε(t)∣α∣ε(t+1)∣=c
Secant Interpolation for Rooting:
In rooting f(x), with initial points x(−1), x(0) set, root of secant interpolation in each step (t) is:
x(t+1)=f(x(t))−f(x(t−1))x(t−1)f(x(t))−x(t)f(x(t−1))
Deduction:
Taylor series of f(x) at x∗ to the second order is:
f(x)=f′(x∗)(x−x∗)+21f′′(x∗)(x−x∗)2
substitute in secant rooting
x(t+1)=f(x(t))−f(x(t−1))x(t−1)f(x(t))−x(t)f(x(t−1))
to obtain
x(t+1)−x∗===f(x(t))−f(x(t−1))x(t−1)f(x(t))−x(t)f(x(t−1))−x∗f′(x∗)(x(t)−x(t−1))−21f′′(x∗)(x(t)−x(t−1))(x(t)+x(t−1)−2x∗)(f′x∗(x(t)−x∗)+21f′′(x∗)(x(t)−x∗)2)(x(t−1)−x∗)−f′(x∗)(x(t)−x(t−1))−21f′′(x∗)(x(t)−x(t−1))(x(t)+x(t−1)−2x∗)(f′x∗(x(t−1)−x∗)+21f′′(x∗)(x(t−1)−x∗)2)(x(t)−x∗)f′(x∗)−21f′′(x∗)(x(t)+x(t−1)−2x∗)21f′′(x∗)(x(t)−x∗)(x(t−1)−x∗)
Denote e(t)≡x(t)−x∗, take t→∞ to obtain
e(t)e(t−1)e(t+1)=2f′(x∗)−f′′(x∗)(e(t)+e(t−1))f′′(x∗)→2f′(x∗)f′′(x∗)
Denote e(t−1)≡a. We could first assume convergence order α and convergence rate c, then
t→∞lim[e(t)]αe(t+1)=t→∞lim[e(t−1)]αe(t)=c⇒e(t+1)=cαaα2,e(t)=caα
substitute into the iteration of error
e(t)e(t−1)e(t+1)=λα−1aα2−α−1=2f′(x∗)f′′(x∗)=const
Considering that limt→∞e(t−1)=0, there must be α2−α−1=0, thus
α=25+1≈1.618