BCSSS

International Encyclopedia of Systems and Cybernetics

2nd Edition, as published by Charles François 2004 Presented by the Bertalanffy Center for the Study of Systems Science Vienna for public access.

About

The International Encyclopedia of Systems and Cybernetics was first edited and published by the system scientist Charles François in 1997. The online version that is provided here was based on the 2nd edition in 2004. It was uploaded and gifted to the center by ASC president Michael Lissack in 2019; the BCSSS purchased the rights for the re-publication of this volume in 200?. In 2018, the original editor expressed his wish to pass on the stewardship over the maintenance and further development of the encyclopedia to the Bertalanffy Center. In the future, the BCSSS seeks to further develop the encyclopedia by open collaboration within the systems sciences. Until the center has found and been able to implement an adequate technical solution for this, the static website is made accessible for the benefit of public scholarship and education.

A B C D E F G H I J K L M N O P Q R S T U V W Y Z

ALGORITHMS (Genetic) 2)

← Back

"A family of methods that search for optimal solutions of difficult problems" (P. DENNING, 1992, p.12)

The concept of genetic algorithm was originally developed by J. HOLLAND as a computer modelization of biological evolution.

P. DENNING writes: "Genetic search algorithms cross-breed trial solutions and allow only the "fittest" solutions (those accorded the highest value) to survive after several generations" (Ibid) They are thus, up to a point, self-perfectible, and a possible model for an evolutive mechanism.

The genetic algorithm avoids at least partly what is possibly the most serious inconvenient of algorithms: their premature stabilization at a sub-optimization level. This result is obtained by a kind of "cooperative competition" by recombination or cross-over between different "candidate" solutions and the introduction during the progressive constructive process of a very slight variability (" mutations") within each elemental situation.

Kenneth DE JONG quoted by P. DENNING, states that "a mutation probability on the order of 0,001 per bit is enough to prevent the search from locking into a local optimum".

In this way, premature sub-optimal solutions are avoided and a global optimum can be more easily reached.

C. EMMECHE resumes as follows the procedure of a genetic algorithm:

"1. Select program pairs on the basis of how well they have solved the task (one can thereby measure fitness). The better the solution, the greater the chance to be selected.

"2. Apply the genetic operator (cross-over, eventually combined with a small chance of mutation) to the selected program pairs in order to create offsprings in the next generation.

"3. Replace the least successful programs with the offspring created in step two, and repeat the process" (1994, p.115)

He adds: "Empirical investigations indicate that this crossing-over scheme operates specially well on problems that programmers otherwise regard as genuinely difficult.

The genetic algorithms are able to exploit the population experience in an optimal manner" (p.116)

→ Parallel distributed processing

Categories

  • 1) General information
  • 2) Methodology or model
  • 3) Epistemology, ontology and semantics
  • 4) Human sciences
  • 5) Discipline oriented

Publisher

Bertalanffy Center for the Study of Systems Science(2020).

To cite this page, please use the following information:

Bertalanffy Center for the Study of Systems Science (2020). Title of the entry. In Charles François (Ed.), International Encyclopedia of Systems and Cybernetics (2). Retrieved from www.systemspedia.org/[full/url]

We thank the following partners for making the open access of this volume possible:

© 2020 BCSSS Contact/Imprint