Conformance Evaluation of Genetic Algorithm for Evolutionary Area Search of Canonical Model
- Authors: Ivanov V.K.1, Palyukh B.V.1, Sotnikov A.N.2
-
Affiliations:
- Tver State Technical University
- Joint Supercomputer Centre
- Issue: Vol 40, No 11 (2019)
- Pages: 1799-1808
- Section: Article
- URL: https://journals.rcsi.science/1995-0802/article/view/206055
- DOI: https://doi.org/10.1134/S1995080219110155
- ID: 206055
Cite item
Abstract
The theory and practice of genetic algorithms is largely based on the Schema Theorem. It was formulated for a canonical genetic algorithm and proves its ability to generate a sufficient number of effective schemata of individuals. Genetic algorithms to solve specific problems and to be different from canonical ones have to be checked to find out whether the Schema Theorem evaluates the algorithm fitness. The article validates the way of testing the algorithm developed as a technique of an area search. The methodology and research results are stated consistently. Coding specifics of the search queries are noted, a criterion of the coding method applicability is substantiated. A variant of the genotype geometric coding is proposed. In comparison with other methods of binary search coding, it provides a short code length and uniqueness as well as conforms the formulated criterion of applicability. Supporting experimental results are given. The Schema Theorem is shown to hold with the iterative execution of the genetic algorithm being tested.
About the authors
V. K. Ivanov
Tver State Technical University
Author for correspondence.
Email: mtivk@tstu.tver.ru
Russian Federation, Tver, 170026
B. V. Palyukh
Tver State Technical University
Author for correspondence.
Email: pboris@tstu.tver.ru
Russian Federation, Tver, 170026
A. N. Sotnikov
Joint Supercomputer Centre
Author for correspondence.
Email: asotnikov@jscc.ru
Russian Federation, Moscow, 119991