Anda di halaman 1dari 2

Newtons Iteration Algorithm: While |r1-r0| > e r0 r1 r1 r0 f(r0)/f(r0) End Needs: f is C2 f has to be available r0 starting point: difficult to find

Theorem: Convergence of Newtons Method F C2 *a,b+, r *a,b+, for which f(r)=0, and f(r)0; then >0, such that r0 [r-,r++ The Newtons sequence will converge to r Secant Method From

Algorithm: Input: f C1 [a,b], tolerance Initialization: r0, r1 in [a,b] While |r1-r0| > r2 r1 f(r1) r1-r0 / f(r1) f(r0) r0 r1 r1 r2

Order of Convergence (of iterative methods) Assume that the sequence converges to r: then if 2 positive constants and such that:

then we say that {rn} converges to r with order and asymptotic error if = 1 ( < 1), the convergence is linear (Bisection, fixed-point) if = 2, the convergence is quadratic (Newton) if 0 < < 1 the convergence is sublinear if 1 < < 2, the convergence is superlinear (Secant = 1.6)

Order of Convergence: Bisection Method Worst case:

Order of Convergence: Fixed-Point Iteration Cn is between rn and r

g(r) = 0 Taylor expansion:

=2

Anda mungkin juga menyukai