Series Prediction Based on Algebraic Approximants
It is described how the Hermite-Padé polynomials corresponding to an algebraic approximant for a power series may be used to predict coefficients of the power series that have not been used to compute the Hermite-Padé polynomials. A recursive algorithm is derived, and some numerical examples are given.
Using sequence transformation and extrapolation algorithms for the prediction of further sequence elements from a finite number of known sequence elements is a topic of growing importance in applied mathematics. For a short introduction, see the book of Brezinski and Redivo Zaglia [1, Section 6.8]. We mention theoretical work on prediction properties of Padé approximants and related algorithms like the epsilon algorithm, and the iterated Aitken and Theta algorithms [2–5], Levin-type sequence transformations [6, 7], the E algorithm [4, 8], and applications on perturbation series of physical problems [7, 9].
Here, we will concentrate on a different class of approximants, namely, the algebraic approximants. For a general introduction to these approximants and the related Hermite-Padé polynomials see . Programs for these approximants are available . We summarize those properties that are important for the following.
Consider a function of complex variable with a known (formal) power series The Hermite-Padé polynomials (HPPs) corresponding to a certain algebraic approximant are polynomials with degree , such that the order condition holds for small . Since one of the coefficients of the polynomials can be normalized to unity, the order condition (1.2) gives rise to a system of linear equations for unknown polynomial coefficients. Thus, the coefficient of of the Taylor expansion at of the left hand side of (1.2) must be zero for . In order to have exactly as many equations as unknowns, we choose and assume that the linear system (1.2) has a solution. Then, the HPPs are uniquely defined upon specifying the normalization. The algebraic approximant under consideration then is that pointwise solution of the algebraic equation for which the Taylor series of coincides with the given power series at least up to order .
We note that for , the algebraic approximants are nothing but the well-known Padé approximants.
Although we assumed that the power series of is known, quite often in practice, only a finite number of coefficients is really known. These coefficients then may be used to compute the Hermite-Padé polynomials and the algebraic approximant under consideration.
We note that the higher coefficients of the Taylor series of may be considered as predictions for the higher coefficients of the power series. The latter are also of interest in applications.
The question then arises how to compute the Taylor series of . If it is possible to solve (1.4) explicitly, that is for , a computer algebra system may be used to do the job. But even then, a recursive algorithm for the computation of the coefficients of the Taylor series would be preferable in order to reduce computational efforts.
In the following section, such a recursive algorithm is obtained. In a further section, we will present numerical examples.
2. The Recursive Algorithm
We consider the HPPs as known. Putting we obtain from (1.4) whence, by equating the coefficient of to zero, we obtain an infinite set of equations. Due to (1.2), all the equations for are satisfied exactly for , .
As a first step, we compute . We note that . Hence, the coefficient of does not involve any terms with . For this coefficient , we only need to consider terms in (2.3) such that , and we obtain for The only terms on the RHS involving are obtained if exactly one of the is equal to , that is, we have , , and for . Thus, we may rewrite all these terms as , where and note that the rest is independent of . Recalling , we obtain Proceeding analogously for , only terms with need to be considered. Hence, for Now, the only terms on the RHS involving are obtained if exactly one of the is equal to , that is, we have , , and for . Thus, we may rewrite all these terms as , where is defined above. Proceeding as before, we put and obtain An equivalent form of the recursive algorithm is obtained in the following way.
Consider for known and the expression It is easy to see, that this expression is exactly equal to , and hence, is linear in the unknown . Thus, we may compute the quantities by substituting into , which entails Thus, starting from , one may compute all the consecutively by repeated use of (2.5), (2.10), and (2.8).
This concludes the derivation of the recursive algorithm.
3. Modes of Application
Basically, there are two modes of application:(a)one computes a sequence of HPPs and for the resulting algebraic approximants, one predicts a fixed number of so far unused coefficients, for example, only one new coefficient. This mode is mainly for tests,(b)one computes from all available coefficients certain HPPs. For the best HPPs one computes a larger number of predictions for so far unused coefficients.
In the following examples, we concentrate on mode (b). Here, it is to be expected that the computed values have the larger errors the higher coefficients are predicted.
The examples serve to introduce to the approach. All numerical calculations in this section were done using Maple (Digits = 16).
Example 4.1. As a first example, we consider , , and, hence, . Since , we are dealing with a quadratic algebraic approximant. Then, the recursive algorithm is started by , . For , we obtain and for , we obtain For the HPPs are determined to be The results for the predicted coefficients given in Table 1.
Example 4.2. As a second example, we consider again , , and , but now the function with the HPPs The results for the predicted coefficients given in Table 2.
Example 4.3. As a final example, we consider the case , whence , and the function The corresponding HPPs are The results for the predicted coefficients are displayed in Table 3.
It is seen that even rather low-order algebraic approximants, or HPPs, respectively, can lead to quite accurate predictions of the unknown coefficients of the power series, especially for , and the next few coefficients.
C. Brezinski and M. Redivo Zaglia, Extrapolation Methods. Theory and Practice, vol. 2 of Studies in Computational Mathematics, North-Holland Publishing Co., Amsterdam, The Netherlands, 1991.
J. Gilewicz, “Numerical detection of the best Padé approximant and determination of the Fourier coefficients of insufficiently sampled functions,” in Padé Approximants and Their Applications, P. R. Graves-Morris, Ed., pp. 99–103, Academic Press, New York, NY, USA, 1973.View at: Google Scholar
E. J. Weniger, “Prediction properties of Aitken's iterated process, of Wynn's epsilon algorithm, and of Brezinski's iterated theta algorithm,” in Numerical Analysis, C. Brezinski, Ed., vol. 2 of Interpolation and Extrapolation, pp. 329–356, Elsevier, Amsterdam, The Netherlands, 2000.View at: Google Scholar
U. D. Jentschura, J. Becher, E. J. Weniger, and G. Soff, “Resummation of QED perturbation series by sequence transformations and the prediction of perturbative coefficients,” Physical Review Letters, vol. 85, no. 12, pp. 2446–2449, 2000.View at: Google Scholar
T. M. Feil and H. H. H. Homeier, “Programs for the approximation of real and imaginary single- and multi-valued functions by means of Hermite-Padé-approximants,” Computer Physics Communications, vol. 158, no. 2, pp. 124–135, 2004, Computer Physics Communications Program Library, Catalogue number: ADSO.View at: Publisher Site | Google Scholar | Zentralblatt MATH