Empirical study of an evaluation-based subdivision algorithm for complex root isolation

  • Authors:
  • Narayan Kamath;Irina Voiculescu;Chee K. Yap

  • Affiliations:
  • Google, Inc. and Oxford Computing Lab, UK;Oxford University, UK;Courant Institute, NYU and Oxford Computing Lab, UK

  • Venue:
  • Proceedings of the 2011 International Workshop on Symbolic-Numeric Computation
  • Year:
  • 2012

Quantified Score

Hi-index 0.00

Visualization

Abstract

We provide an empirical study of subdivision algorithms for isolating the simple roots of a polynomial in any desired box region B0 of the complex plane. One such class of algorithms is based on Newton-like interval methods (Moore, Krawczyk, Hansen-Sengupta). Another class of subdivision algorithms is based on function evaluation. Here, Yakoubsohn discussed a method that is purely based on an exclusion predicate. Recently, Sagraloff and Yap introduced another algorithm of this type, called Ceval. We describe the first implementation of Ceval in Core Library. We compare its performance to the above mentioned algorithms, and also to the well-known MPSolve software from Bini and Florentino. Our results suggest that certified evaluation-based methods such as Ceval are encouraging and deserve further exploration.