Content-Length: 151261 | pFad | https://dlmf.nist.gov/./.././not/.././bib/../././././3.9#Px1.info
All sequences (series) in this section are sequences (series) of real or complex numbers.
A transformation of a convergent sequence with limit into a sequence is called limit-preserving if converges to the same limit .
The transformation is accelerating if it is limit-preserving and if
3.9.1 | |||
Similarly for convergent series if we regard the sum as the limit of the sequence of partial sums.
It should be borne in mind that a sequence (series) transformation can be effective for one type of sequence (series) but may not accelerate convergence for another type. It may even fail altogether by not being limit-preserving.
If is a convergent series, then
3.9.2 | |||
provided that the right-hand side converges. Here is the forward difference operator:
3.9.3 | |||
. | |||
Thus
3.9.4 | |||
Euler’s transformation is usually applied to alternating series. Examples are provided by the following analytic transformations of slowly-convergent series into rapidly convergent ones:
3.9.5 | |||
3.9.6 | |||
3.9.7 | |||
This transformation is accelerating if is a linearly convergent sequence, i.e., a sequence for which
3.9.8 | |||
. | |||
When applied repeatedly, Aitken’s process is known as the iterated -process. See Brezinski and Redivo Zaglia (1991, pp. 39–42).
Shanks’ transformation is a generalization of Aitken’s -process. Let be a fixed positive integer. Then the transformation of the sequence into a sequence is given by
3.9.9 | |||
, | |||
where is the Hankel determinant
3.9.10 | |||
The ratio of the Hankel determinants in (3.9.9) can be computed recursively by Wynn’s epsilon algorithm:
3.9.11 | ||||
, | ||||
. | ||||
Then . Aitken’s -process is the case .
If is the th partial sum of a power series , then is the Padé approximant (§3.11(iv)).
For further information on the epsilon algorithm see Brezinski and Redivo Zaglia (1991, pp. 78–95).
0 | 0.80000 00000 00 | 0.82182 62806 24 | 0.82244 84501 47 | 0.82246 64909 60 | 0.82246 70175 41 |
---|---|---|---|---|---|
1 | 0.82692 30769 23 | 0.82259 02017 65 | 0.82247 05346 57 | 0.82246 71342 06 | 0.82246 70363 45 |
2 | 0.82111 11111 11 | 0.82243 44785 14 | 0.82246 61821 45 | 0.82246 70102 48 | 0.82246 70327 79 |
3 | 0.82300 13550 14 | 0.82247 78118 35 | 0.82246 72851 83 | 0.82246 70397 56 | 0.82246 70335 90 |
4 | 0.82221 76684 88 | 0.82246 28314 41 | 0.82246 69467 93 | 0.82246 70314 36 | 0.82246 70333 75 |
5 | 0.82259 80392 16 | 0.82246 88857 22 | 0.82246 70670 21 | 0.82246 70341 24 | 0.82246 70334 40 |
6 | 0.82239 19390 77 | 0.82246 61352 37 | 0.82246 70190 76 | 0.82246 70331 54 | 0.82246 70334 18 |
7 | 0.82251 30483 23 | 0.82246 75033 13 | 0.82246 70400 56 | 0.82246 70335 37 | 0.82246 70334 26 |
8 | 0.82243 73137 33 | 0.82246 67719 32 | 0.82246 70301 49 | 0.82246 70333 73 | 0.82246 70334 23 |
9 | 0.82248 70624 89 | 0.82246 71865 91 | 0.82246 70351 34 | 0.82246 70334 48 | 0.82246 70334 24 |
10 | 0.82245 30535 15 | 0.82246 69397 57 | 0.82246 70324 88 | 0.82246 70334 12 | 0.82246 70334 24 |
We give a special form of Levin’s transformation in which the sequence of partial sums is transformed into:
3.9.13 | |||
where is a fixed nonnegative integer, and
3.9.14 | |||
Sequences that are accelerated by Levin’s transformation include logarithmically convergent sequences, i.e., sequences converging to such that
3.9.15 | |||
For further information see Brezinski and Redivo Zaglia (1991, pp. 39–42).
Fetched URL: https://dlmf.nist.gov/./.././not/.././bib/../././././3.9#Px1.info
Alternative Proxies: