Metadata language
Biologically inspired methods for control of evolutionary algorithms
Subtitle:Raport Badawczy = Research Report ; RB/89/2003
Creator: Publisher:Instytut Badań Systemowych. Polska Akademia Nauk ; Systems Research Institute. Polish Academy of Sciences
Place of publishing: Date issued/created: Description:[411]- 433 pages ; 21 cm ; Bibliography p. 432-433
Abstract:In this paper two methods for evolutionary algorithm control are proposed. The first one is a new method of tuning the probabilities of genetic operators. It is assumed in the presented approach that every member of the optimized population conducts his own ranking of genetic operators' qualities. This ranking enables computing the probabilities of execution of genetic operators. This set of probabilities is a basis of experience of every individual and according to this basis the individual chooses the operator in every iteration of the algorithm. Due to this experience one can maximize the chances of his offspring to survive. The second part of the paper deals with a self-adapting method of selection of individuals to a subsequent generation. Methods of selection applied in the evolutionary algorithms are usually inspired by nature and prefer solutions where the main role is played by randomness, competition and struggle among individuals. In the case of evolutionary algorithms, where populations of individuals are usually small, this causes a premature convergence to local minima.
Relation:Raport Badawczy = Research Report
Resource type: Detailed Resource Type: Source: Language: Language of abstract: Rights:Creative Commons Attribution BY 4.0 license
Terms of use:Copyright-protected material. [CC BY 4.0] May be used within the scope specified in Creative Commons Attribution BY 4.0 license, full text available at: ; -
Digitizing institution:Systems Research Institute of the Polish Academy of Sciences
Original in:Library of Systems Research Institute PAS
Projects co-financed by: Access: