EUVIMEDEUVIMED entdecken
Crossref

Energy efficiency and performance of automated search for minmax approximation with guaranteed accuracy in finite precision

Andrei A. Chusov, Anna A. Shklyar, Lyubov G. Statsenko, Alexander Yu. Rodionov

Vestnik of Saint Petersburg University. Applied Mathematics. Computer Science. Control Processes · 2025 · 21 (3) · 401-416

DOI: 10.21638/spbu10.2025.307 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Vestnik of Saint Petersburg University. Applied Mathematics. Computer Science. Control Processes, 2025. Verfasst von Andrei A. Chusov und 3 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The paper describes proposed results of formal analysis of errors resulting from parallel search for minimax polynomials approximating functions in finite numerical precision, a method of parallel search for the search and analysis of its performance in terms of time and energy consumption. This analysis is required because often the known axioms of algebraic fields are erroneously assumed true in the cases of finite precision computer arithmetics, including finite rational Euclidean spaces and floating-point systems. On the other hand, these axioms are required in order to rearrange computations to achieve greater speeds and energy efficiency of the compound computation. For instance, such is the case of parallel computations addressed in the paper. Well-known generalizations of the Weierstrass approximation theorem to the cases of complete metric spaces with maps from the corresponding Banach space provide for implementation of the Remez algorithm for search of the minimax polynomials to the case of finite rational spaces, including finite-precision numerical systems. This version of the Remez algorithm, based on the Householder’s numerical method of solving systems of equations, is described in the paper in the context of parallel computing and its effect on approximation errors in finite numerical precision as well as time performance and energy efficiency. The results presented in the paper agree with the results of the experiments whose implementation is available to reader for verification as addressed in the paper.

Zitieren

APA

Chusov, A. A., Shklyar, A. A., Statsenko, L. G., & Rodionov, A. Y. (2025). Energy efficiency and performance of automated search for minmax approximation with guaranteed accuracy in finite precision. Vestnik of Saint Petersburg University. Applied Mathematics. Computer Science. Control Processes, 21(3), 401-416. https://doi.org/10.21638/spbu10.2025.307

Vancouver

Chusov AA, Shklyar AA, Statsenko LG, Rodionov AY. Energy efficiency and performance of automated search for minmax approximation with guaranteed accuracy in finite precision. Vestnik of Saint Petersburg University. Applied Mathematics. Computer Science. Control Processes. 2025;21(3):401-416. doi:10.21638/spbu10.2025.307

BibTeX
@article{Chusov2025,
  author = {Chusov, Andrei A. and Shklyar, Anna A. and Statsenko, Lyubov G. and Rodionov, Alexander Yu.},
  title = {Energy efficiency and performance of automated search for minmax approximation with guaranteed accuracy in finite precision},
  journal = {Vestnik of Saint Petersburg University. Applied Mathematics. Computer Science. Control Processes},
  year = {2025},
  volume = {21},
  number = {3},
  pages = {401--416},
  doi = {10.21638/spbu10.2025.307}
}
Kostenlos starten →