TY - GEN
T1 - Novel Range Functions via Taylor Expansions and Recursive Lagrange Interpolation with Application to Real Root Isolation
AU - Hormann, Kai
AU - Kania, Lucas
AU - Yap, Chee
N1 - Funding Information:
Chee’s work began under a USI Visiting Professor Fellowship (Fall 2018); additional support comes from NSF Grants # CCF-1564132 and # CCF-2008768.
Funding Information:
Lucas was supported by a UROP Fellowship (Summer 2019) from the Faculty of Informatics at Università della Svizzera italiana (USI).
Funding Information:
Lucas was supported by a UROP Fellowship (Summer 2019) from the Faculty of Informatics at Universit? della Svizzera italiana (USI). Chee's work began under a USI Visiting Professor Fellowship (Fall 2018); additional support comes from NSF Grants # CCF-1564132 and # CCF-2008768.
Publisher Copyright:
© 2021 ACM.
PY - 2021/7/18
Y1 - 2021/7/18
N2 - Range functions are an important tool for interval computations, and they can be employed for the problem of root isolation. In this paper, we first introduce two new classes of range functions for real functions. They are based on the remainder form by Cornelius and Lohner [7] and provide different improvements for the remainder part of this form. On the one hand, we use centered Taylor expansions to derive a generalization of the classical Taylor form with higher than quadratic convergence. On the other hand, we propose a recursive interpolation procedure, in particular based on quadratic Lagrange interpolation, leading to recursive Lagrange forms with cubic and quartic convergence. We then use these forms for isolating the real roots of square-free polynomials with the algorithm Eval, a relatively recent algorithm that has been shown to be effective and practical. Finally, we compare the performance of our new range functions against the standard Taylor form. Range functions are often compared in isolation; in contrast, our holistic comparison is based on their performance in an application. Specifically, Eval can exploit features of our recursive Lagrange forms which are not found in range functions based on Taylor expansion. Experimentally, this yields at least a twofold speedup in Eval.
AB - Range functions are an important tool for interval computations, and they can be employed for the problem of root isolation. In this paper, we first introduce two new classes of range functions for real functions. They are based on the remainder form by Cornelius and Lohner [7] and provide different improvements for the remainder part of this form. On the one hand, we use centered Taylor expansions to derive a generalization of the classical Taylor form with higher than quadratic convergence. On the other hand, we propose a recursive interpolation procedure, in particular based on quadratic Lagrange interpolation, leading to recursive Lagrange forms with cubic and quartic convergence. We then use these forms for isolating the real roots of square-free polynomials with the algorithm Eval, a relatively recent algorithm that has been shown to be effective and practical. Finally, we compare the performance of our new range functions against the standard Taylor form. Range functions are often compared in isolation; in contrast, our holistic comparison is based on their performance in an application. Specifically, Eval can exploit features of our recursive Lagrange forms which are not found in range functions based on Taylor expansion. Experimentally, this yields at least a twofold speedup in Eval.
KW - eval algorithm
KW - interval arithmetic
KW - lagrange form
KW - range functions
KW - root isolation
KW - taylor form
UR - http://www.scopus.com/inward/record.url?scp=85111113452&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85111113452&partnerID=8YFLogxK
U2 - 10.1145/3452143.3465532
DO - 10.1145/3452143.3465532
M3 - Conference contribution
AN - SCOPUS:85111113452
T3 - Proceedings of the International Symposium on Symbolic and Algebraic Computation, ISSAC
SP - 193
EP - 200
BT - ISSAC 2021 - Proceedings of the 2021 International Symposium on Symbolic and Algebraic Computation
PB - Association for Computing Machinery
T2 - 46th International Symposium on Symbolic and Algebraic Computation, ISSAC 2021
Y2 - 18 July 2021 through 23 July 2021
ER -