Signal and noise: Towards a general theory of algorithms

Author(s): Jones, B., Iyengar, S.

Journal/Book: Cybernetica. 1997; 40: Palais Des Expositions, Place Andre Rijckmans, 5000 Namur, Belgium. Int Assn Cybernetics. 151-164.

Abstract: This paper presents a generalization of iterative numerical algorithms. An algorithm is considered to be composed of principal parts. Time series are associated with the algorithm and each of its principal parts. Each time series breaks down into dynamic signal and noise components. The problem treated in this immediate work is to extract signal from noise in the components of slow moving algorithms; thereby rapidly obtaining a solution. A program has been written to do extraction for slow algorithms, and it is applied to the principal parts of a well known algorithm for a well known characteristic test problem. The framework presented is general for analyzing and improving iterative numerical algorithms.

Note: Article Trivedi S, Southern Univ, Dept Comp Sci, Baton Rouge,LA 70813 USA

Keyword(s): algorithm theory; signal; noise; extrapolation

Zurück | Weiter