SEQUENCE_TRANSFORMATIONS

(Redirected from Aitken\'s delta-squared process)
In mathematics, a 'sequence transformation' is a resummation of a sequence. Sequence transformations are commonly used for series acceleration, that is, for improving the rate of convergence of a slowly convergent sequence or series. Sequence transformations are also commonly used to compute the antilimit of a divergent series numerically, and are used in conjunction with extrapolation methods.

Contents

Overview

Definitions

Application

Example: The Aitken method (Aitken extrapolation)

Example: minimum polynomial extrapolation

Example: Euler's transform

References

Overview

Techniques for series acceleration are often applied in numerical analysis, where they are used to improve the speed of numerical integration. Series acceleration techniques may also be used, for example, to obtain a variety of identities on special functions. Thus, the Euler transform applied to the hypergeometric series gives some of the classic, well-known hypergeometric series identities.
Two classical techniques for series acceleration are Euler's transformation of series^[1] and Kummer's transformation of series^[1]. A variety of much more rapidly convergent and special-case tools have been developed in the 20th century, including Richardson extrapolation, introduced by Lewis Fry Richardson in the early 20th century but also known and used by Hirotaka Takebe in 1722, the Aitken delta-squared process, introduced by Alexander Aitken in 1926 but also known and used by Takakazu Seki in the 18th century, the e-algorithm given by Peter Wynn in 1956, the Levin u-transform, and the Wilf-Zeilberger-Ekhad method or WZ method.
For alternating series, several powerful techniques, offering convergence rates from $5.828^\{-n\}$ all the way to $17.93^\{-n\}$ for a summation of $n$ terms, are described by Cohen ''et al''. ^[3].

Definitions

For a given sequence
:$S=\{\; s\_n\; \}\_\{nin\; N\_0\}$,
the 'transformed sequence' $mathbb\{T\}$ is
:$mathbb\{T\}(S)=S\text{'}=\{\; s\text{'}\_n\; \}\_\{nin\; N\_0\}$,
where the members of the transformed sequence are usually computed from some finite number of members of the original sequence, e.g., there is a mapping $T$ according to
:$T\; :\; left(s\_n,s\_\{n+1\},dots,s\_\{n+k\}$
ight) o s'_n
for some finite $k$. In the simplest case, the $s\_n$ and the $s\text{'}\_n$ are real or complex numbers. More generally, they may be elements of some vector space or algebra.
The transformed sequence is said to 'converge faster' than the original sequence if
:
where $s$ is the limit of $S$ assumed to be convergent. In this case, convergence acceleration is obtained. If the original sequence is divergent, the sequence transformation acts as extrapolation method to the antilimit $s$.
If the mapping $T$ is linear in each of its arguments, i.e., for
:$s\text{'}\_n=sum\_\{m=0\}^\{k\}\; c\_m\; s\_\{n+m\}$
for some constants $c\_0,dots,c\_k$, the sequence transformation $mathbb\{T\}$ is called a 'linear sequence transformation'. Sequence transformations that are not linear are called nonlinear sequence transformations.
Classical examples of linear sequence transformations are Euler's transformation of series, from the 18th century, and Kummer's transformation of series^[1].

Application

Examples of such nonlinear sequence transformations are Padé approximants and Levin-type sequence transformations.
Especially nonlinear sequence transformations often provide powerful numerical methods for the summation of divergent series or asymptotic series that arise for instance in perturbation theory, and may be used as highly effective extrapolation methods.

Example: The Aitken method (Aitken extrapolation)

A simple nonlinear sequence transformation is the 'Aitken method'
:$mathbb\{A\}\; :\; S\; o\; S\text{'}=mathbb\{A\}(S)$
defined by the mapping
:$A\; :\; left(s\_n,s\_\{n+1\},s\_\{n+2\}$
ight) o s'_n
with
:.
Taking the sequence of partial sums
:
of the geometric series as untransformed sequence, we obtain for the transformed sequence by simple algebra
:
independent of $n$. This, however is the limit of the geometric series for 1=|''q''| < 1. Thus, the Aitken method yields the exact result by extrapolation of only three consecutive partial sums. For 1=|''q''| > 1, the geometric series diverges since it is a power series in ''q'' outside its radius of convergence. But even for this divergent series, the Aitken method sums the geometric series to its analytic continuation for all complex numbers ''q'' ≠ 1.
The Aitken method is often used iteratively by applying it again to the transformed sequence ''S''′:
:$displaystyle\; S\text{'}\text{'}=mathbb\{A\}(S\text{'})$
:$displaystyle\; S\text{'}=mathbb\{A\}(S\text{'}\text{'})$
and so on.

Example: minimum polynomial extrapolation

While Aitken's method is the most famous, it often fails for vector sequences. An effective method for vector sequences is the Minimum Polynomial Extrapolation. It is usually phrased in terms of the fixed point iteration:
: $x\_\{k+1\}=f(x\_k).$
Given iterates $x\_1,\; x\_2,\; ...,\; x\_k$ in $Bbb\; R^n$, one constructs the $n\; imes\; (k-1)$ matrix $U=(x\_2-x\_1,\; x\_3-x\_2,\; ...,\; x\_k-x\_\{k-1\})$ whose columns are the $k-1$ differences. Then, one computes the vector $c=U^+\; (x\_\{k+1\}-x\_k)$ where $U^+$ denotes the Moore-Penrose Pseudoinverse of $U$. The number 1 is then appended to the end of $c$, and the extrapolated limit is
:$s=\{\; X\; c\; over\; sum\_\{i=1\}^k\; c\_i\; \},$
where $X=(x\_2,\; x\_3,\; ...,\; x\_\{k+1\})$ is the matrix whose columns are the $k$ iterates starting at 2.
The following 4 line MATLAB code segment implements the MPE algorithm:

U=x(:,2:end-1)-x(:,1:end-2);

c=-pinv(U)
★ (x(:,end)-x(:,end-1));

c(end+1,1)=1;

s=(x(:,2:end)
★ c)/sum(c);

Example: Euler's transform

For example, Euler's transform is given by
:$sum\_\{n=0\}^infty\; (-1)^n\; a\_n\; =\; sum\_\{n=0\}^infty\; (-1)^n$
rac {Delta^n a_0} {2^{n+1}}
where $Delta$ is the forward difference operator:
:$Delta^n\; a\_0\; =\; sum\_\{k=0\}^n\; (-1)^k\; \{n\; choose\; k\}\; a\_\{n-k\}$
If the original series, on the left hand side, is only slowly converging, the forward differences will tend to become small quite rapidly; the additional power of two further improves the rate at which the left hand side converges.
A particularly efficient numerical implementation of the Euler transform is the van Wijngaarden transformation^[5]

References

1.
2.
3. Henri Cohen, Fernando Rodriguez Villegas, and Don Zagier,
"Convergence Acceleration of Alternating Series", ''Experimental Mathematics'', '9':1 (2000) page 3.
4.
5. William H. Press, ''et al'', ''Numerical Recipes in C'', (1987) Cambridge University Press, ISBN 0-521-43108-5 ''(See section 5.1)''


★ C. Brezinski and M. Redivo Zaglia, ''Extrapolation Methods. Theory and Practice'', North-Holland, 1991.

★ G. A. Baker, Jr. and P. Graves-Morris, ''Padé Approximants'', Cambridge U.P., 1996.