Publication detail
Optimizing GA Lifetime Parameters analyzing Error Probability Estimates
ROUPEC, J. POPELA, P. OŠMERA, P.
Czech title
Optimalizace doby života jedince v GA použitím odhadu pravděpodobnosti chyby
English title
Optimizing GA Lifetime Parameters analyzing Error Probability Estimates
Type
conference paper
Language
en
Original abstract
The purpose of the paper is to discuss the qualitative properties of a certain class of genetic algorithms from the reliability viewpoint. The reliability of the algorithm is understood as the seldom occurance of big computational errors. This concept is measured either by a simple error probability estimate for a given error level or by a error level estimate for a given probability. Then the suitable value of the lifetime parameter is chosen minimizing above-mentioned measures. A short remark is addressed to the stability of such choice. General ideas are tested for a specific GA class with an advanced crossover operator. It also includes a lifetime limit of population element. All computations are realized for an example of staircase function identification.
Czech abstract
Účelem článku je diskutovat kvalitu jisté třídy GA z hlediska spolehlivosti. Nespolehlivostí algoritmu se rozumí občasný výskyt velkých výpočetních chyb. Tento koncept je vyčíslen buď jednoduchou pravděpodobností chyby nebo odhadem chyby pro danou pravděpodobnost. Následně je zvolena vhodná hodnota parametru doby života tak, aby minimalizovala výše zmíněné hodnoty. Krátce je diskutována i stabilita této volby.
English abstract
The purpose of the paper is to discuss the qualitative properties of a certain class of genetic algorithms from the reliability viewpoint. The reliability of the algorithm is understood as the seldom occurance of big computational errors. This concept is measured either by a simple error probability estimate for a given error level or by a error level estimate for a given probability. Then the suitable value of the lifetime parameter is chosen minimizing above-mentioned measures. A short remark is addressed to the stability of such choice. General ideas are tested for a specific GA class with an advanced crossover operator. It also includes a lifetime limit of population element. All computations are realized for an example of staircase function identification.
Keywords in English
genetic algorithm, lifetime limit, adaptability, dynamic environment, diploid chromosomes, shades, error probability, probability and quantile minimization
RIV year
2000
Released
07.06.2001
Publisher
PC-DIR
Location
Brno, Czech Republic
ISBN
80-214-1609-2
Book
The 6th International Conference on Soft Computing MENDEL 2000
Pages count
6
BIBTEX
@inproceedings{BUT2218,
author="Jan {Roupec} and Pavel {Popela} and Pavel {Ošmera},
title="Optimizing GA Lifetime Parameters analyzing Error Probability Estimates",
booktitle="The 6th International Conference on Soft Computing MENDEL 2000",
year="2001",
month="June",
publisher="PC-DIR",
address="Brno, Czech Republic",
isbn="80-214-1609-2"
}