Cross-Selection Based Evolution Strategies

Abstract

A search for an optimal value of a complex multi-dimensional continuous function is still one of the most pressing problems. The genetic algorithms (GA) and evolution strategies (ES) are methods to solving optimization problems that is based on natural selection, the process that drives biological evolution. Our goal was to use evolutionary optimization methods to find the global optimal value (minimum) of a non-smooth multi-dimensional function with a large number of local minimums. We took several test functions of different levels of complexity and used evolution strategies to solve the problem. The standard evolution strategies, which work well with smooth functions, gave us various points of local minimums as a solution, without finding the global minimum, for the complex function. In our work, we propose a new approach: the cross-selection method, which, in combination with previously developed methods - adaptive evolution strategies, gave a good result for the searth for the global minimum the complex function.