Algorithms for simultaneous Hermite–Padé approximations

Johan Rosenkilde, Arne Storjohann

Research output: Contribution to journalJournal articleResearchpeer-review

12 Downloads (Pure)

Abstract

We describe how to compute simultaneous Hermite–Padé approximations, over a polynomial ring K[x] for a field K using O(nω−1td) operations in K, where d is the sought precision, where n is the number of simultaneous approximations using t<n polynomials, and where O(nω) is the cost of multiplying n×n matrices over K. We develop two algorithms using different approaches. Both algorithms return a reduced sub-basis that generates the complete set of solutions to the input approximation problem that satisfy the given degree constraints. Previously, the cost O(nω−1td) has only been reached with randomized algorithms finding a single solution for the case t<n. Our results are made possible by recent breakthroughs in fast computations of minimal approximant bases and Hermite–Padé approximations for the case t≥n.

Original languageEnglish
JournalJournal of Symbolic Computation
Volume102
Pages (from-to)279-303
ISSN0747-7171
DOIs
Publication statusPublished - 1 Jan 2020

Keywords

  • Approximant bases
  • Padé approximation
  • Structured linear systems

Fingerprint Dive into the research topics of 'Algorithms for simultaneous Hermite–Padé approximations'. Together they form a unique fingerprint.

Cite this