Power series remainder sequences and Padé fractions over an integral domain

  • Authors:
  • S. Cabay;P. Kossowski

  • Affiliations:
  • Department of Computing Science, University of Alberta, Edmonton, Alberta, T6G 2H1, Canada;Department of Mathematics, University of Kuwait, Kuwait

  • Venue:
  • Journal of Symbolic Computation
  • Year:
  • 1990

Quantified Score

Hi-index 0.03

Visualization

Abstract

For univariate power series with coefficients over an integral domain, the notions of power series pseudo-division and power series remainder sequences are introduced and developed. By appealing to the theory of resultants, an algorithm is developed for constructing power series remainder sequences in which the sizes of the coefficients of successive remainders grow linearly, only. The algorithm computes also a co-factor sequence, which is shown to be associated with a given power series remainder sequence. The co-factor sequence yields directly all the Pade fractions along some specific off-diagonal path of the Pade table for a pair of power series (A(z), B(z)). When the sizes of the coefficients of A(z) and B(z) are bounded by k, in the normal case, the algorithm computes Pade fractions of type (m,n) in time O(k^2(m+n)^4). In the abnormal case, and depending on the nature of abnormalities, the time complexity can reach O(k^2(m+n)^5). For computing Pade fractions, the algorithm compares favorably with fraction-free methods, which have a time complexity of O(k^2(m+n)^5), even in the normal case. When applied to polynomials, rather than to power series, the algorithm, for one specific off-diagonal path, corresponds to Euclid's extended algorithm for computing the greatest common divisor of two polynomials.