Article 10319

Title of the article

ANALYSIS OF THE PROCESS OF EVOLUTIONARY GENETIC CALCULATIONS IN TERMS OF GENERALIZATION CHARACTERISTICS 

Authors

Imamutdinov Anton Igorevich, postgraduate student, Penza State University (440026, 40 Krasnaya street, Penza, Russia), E-mail: antonim94@yandex.ru
Sleptsov Nikolay Vladimirovich, candidate of technical sciences, associate professor, sub-department of economic cybernetics, Penza State University (440026, 40 Krasnaya street, Penza, Russia), E-mail: nbs_nbs@km.ru 

Index UDK

519.718 

DOI

10.21685/2307-4205-2019-3-10 

Abstract

Background. Evolutionary – genetic calculations are actively used to solve complex problems, regardless of the formal – logical characteristics of the relationships between the parameters of these problems. Their main advantage consists in the possibility of obtaining some coding of the solution based on minimal information about the input data arrays and the response of the model to the input data. Expanding the application of methods requires, among other tasks of ensuring high indicators of generalization for the solution obtained.
Methods. Consideration of the properties of the decisionmaking process in evolutionary computations is carried out using the example of finding a strategy for solving the DZ game problem – the dilemma of the prisoner, belonging to the class of noncooperative games with a nonzero sum and widely used in game theory and machine learning.
Results. The study of the model revealed the reasons for the formation of the most effective behavioral strategies and features imposed by the evolutionary nature of the result.
Conclusions. The possibility of increasing the generalization indices for solutions obtained by the methods of evolutionary computing using the proposed approaches has been proved. 

Key words

evolutionary computations, genetic algorithms, training, game strategies, mass exclusions, superspecialization, convergence 

Download PDF

 

Дата создания: 01.11.2019 13:02
Дата обновления: 01.11.2019 14:05