Procedure of Selecting Stochastic Search Multi-agent Algorithm Parameters by the Example of Solving Discrete Optimization Problems
Oleg Yesikov, D.O. Yesikov, В. Ю. Румянцев, Alexey N. Ivutin · 2019
A list of practical discrete optimization problems, which due to their high dimensionality, nonlinear nature of objective functions and restrictions cannot be solved by means of conventional methods, is presented. Application of stochastic search multi-agent algorithms with island computing organization scheme for solving indicated problems is validated experimentally. Multi-agent algorithm parameters found experimentally and approximating dependences to determine them, which in solving problems of various dimensionalities provide obtaining the highest possible quality of result within restricted time, are presented. The possibility is verified experimentally, and mathematical dependences, the use of which enables controlling accuracy and time of solving by means of algorithm parameter variation, are defined.