Научная электронная библиотека ULRICH'S Periodicals Directory

Сибирский государственный университет
науки и технологий имени академика М.Ф. Решетнева

Сибирский аэрокосмический журнал
ISSN 2712-8970 (Print) ISSN 2782-5760 (on-line)

Сведения об образовательной организации

UDK UDC 519.6 Doi: 10.31772/2587-6066-2019-20-4-436-442
COMPARISON OF METHODS FOR INITIALIZING STARTING POINTS ON THE OPTIMIZATION GENETIC ALGORITHM
Pavlenko A. A.
Reshetnev Siberian State University of Science and Technology; 31, Krasnoyarsky Rabochy Av., Krasnoyarsk, 660037, Russian Federation; E-mail: saaprepod@mail.ru
The way to initialize the starting points for optimization algorithms is one of the main parameters. Currently used methods of initializing starting points are based on stochastic algorithms of spreading points. In a genetic algorithm, points are Boolean sets. These lines are formed in different ways. They are formed directly, using random sequences (with uniform distribution law) or formed using random sequences (with uniform distribution law) in the space of real numbers, and then converted to boolean numbers. Six algorithms for constructing multidimensional points for global optimization algorithms of boolean sets based on both stochastic and non-random point spreading algorithms are designed. The first four methods of initialization of Boolean lines used a random distribution law, and the fifth and sixth methods of initialization used a non-random method of forming starting points-LP sequence. A large number of optimization algorithms were restarted. Calculations of high accuracy were used. The research was carried out on the genetic algorithm of global optimization. The work is based on Acly function, Rastrigin function, Shekel function, Griewank function and Rosenbrock function. The research was based on three algorithms of srarting points spreading: LP sequence, UDC sequence, regular random spreading. The best parameters of the genetic algorithm of global optimization were used in the work. As a result, we obtained arrays of mathematical expectations and standard deviations of the solution quality for different functions and optimization algorithms. The purpose of the analysis of ways to initialize the starting points for the genetic optimization algorithm was to find the extremum quickly, accurately, cheaply and reliably simultaneously. Methods of initialization were compared with each other by expectation and standard deviation. The quality of the solution is understood as the average error of finding the extremum. The best way of initialization of starting points for genetic optimization algorithm on these test functions is revealed.
Keywords: genetic optimization algorithm, points initialization methods.
References

1. Zaloga A. N., Yakimov I. S., Dubinin P. S. Multipopulation Genetic Algorithm for Determining Crystal
Structures Using Powder Diffraction Data. Journal of Surface Investigation: X-ray, Synchrotron and Neutron
Techniques. 2018, Vol. 12, No. 1, P. 128–134.
2. Stanovov V., Akhmedova S., Semenkin E. Automatic Design of Fuzzy Controller for Rotary Inverted
Pendulum with Success-History Adaptive Genetic Algorithm. 2019 International Conference on Information
Technologies (InfoTech). IEEE, 2019. P. 1–4.
3. Zaloga A. et al. Genetic Algorithm for Automated X-Ray Diffraction Full-Profile Analysis of Electrolyte
Composition on Aluminium Smelters. Informatics in Control, Automation and Robotics 12th International
Conference, ICINCO 2015 Colmar, France, July 21–23, 2015 Revised Selected Papers. Springer, Cham, 2016,
P. 79–93.
4. Du X. et al. Genetic algorithm optimized nondestructive prediction on property of mechanically injured
peaches during postharvest storage by portable visible/shortwave near-infrared spectroscopy. Scientia
Horticulturae. 2019, Vol. 249, P. 240–249.
5. Akhmedova S., Stanovov V., Semenkin E. Soft Island Model for Population-Based Optimization Algorithms.
International Conference on Swarm Intelligence. Springer, Cham, 2018, P. 68–77.
6. Yakimov I. et al. Application of Evolutionary Rietveld Method Based XRD Phase Analysis and a Self-
Configuring Genetic Algorithm to the Inspection of Electrolyte Composition in Aluminum Electrolysis Baths.
Crystals. 2018, Vol. 8, No. 11, P. 402.
7. Chen W. et al. Applying population-based evolutionary algorithms and a neuro-fuzzy system for
modeling landslide susceptibility. Catena. 2019, Vol. 172, P. 212–231.
8. Karabadzhak G. et al. Semi-Empirical Method for Evaluation of a Xenon Operating Hall Thruster Erosion
Rate Through Analysis of its Emission Spectra. Spacecraft Propulsion. 2000, Vol. 465, P. 909.
9. Akhmedova S., Stanovov V., Semenkin E. Success-History Based Position Adaptation in Co-operation of
Biology Related Algorithms. International Conference on Swarm Intelligence. Springer, Cham, 2019, P. 39–49.
10. Yakimov I., Zaloga A., Dubinin P., Bezrukovа O., Samoilo A., Burakov S., Semenkin E., Semenkina M.,
Andruschenko E. Application of Evolutionary Rietveld Method Based XRD Phase Analysis and a Self-
Configuring Genetic Algorithm to the Inspection of Electrolyte Composition in Aluminum Electrolysis
Baths. Crystals. 2018, Vol. 8, No. 11, P. 402.
11.Brester C., Rönkkö M., Kolehmainen M., Semenkin E., Kauhanen J., Tuomainen T. P., Voutilainen S.,
Ronkainen K. Evolutionary methods for variable selection in the epidemiological modeling of cardiovascular
diseases. BioData Mining. 2018, Vol. 11, No. 18.
12.Mamontov D. et al. Evolutionary Algorithms for the Design of Neural Network Classifiers for the Classification
of Pain Intensity. IAPR Workshop on Multimodal Pattern Recognition of Social Signals in Human-
Computer Interaction. Springer, Cham, 2018, P. 84–100.
13. Brester C. et al. On a restart metaheuristic for realvalued multi-objective evolutionary algorithms. Proceedings
of the Genetic and Evolutionary Computation Conference Companion. ACM, 2019, P. 197–198.
14. Sharifpour E. et al. Zinc oxide nanorod loaded activated carbon for ultrasound assisted adsorption of safranin
O: Central composite design and genetic algorithm optimization Applied Organometallic Chemistry. 2018,
Vol. 32, No. 2, P. e4099.
15. Penenko A. V. 2019 Newton–Kantorovich method for solving inverse problems of source identification
in product–destruction models with time series data. Siberian J. of computational mathematics. 2019, No. 1,
P. 57–79.


Pavlenko Alexander Alexandrovich – Senior Lecturer; Reshetnev Siberian State University of Science and
Technology. E-mail: saaprepod@mail.ru.


  COMPARISON OF METHODS FOR INITIALIZING STARTING POINTS ON THE OPTIMIZATION GENETIC ALGORITHM