global optimization;controlled random search;competing heuristics; adaptable algorithms

The goal of the paper is to study stochastic algorithms with high self-adaptation which could be used in "routine" task when the global optimization problem is to be solved. A~generalization of the controlled random search (CRS) algorithm is described. Several heuristics generating a new trial point are used at random with probabilities depending on the success of heuristics in preceding steps. This generalized CRS algorithm was compared experimentally with the differential evolution on several test functions. The results showed that the proposed algorithm was more reliable and more effective in most of the test functions. Moreover, the proposed algorithm does not need a special tuning of its input parameters for each task.

The goal of the paper is to study stochastic algorithms with high self-adaptation which could be used in "routine" task when the global optimization problem is to be solved. A~generalization of the controlled random search (CRS) algorithm is described. Several heuristics generating a new trial point are used at random with probabilities depending on the success of heuristics in preceding steps. This generalized CRS algorithm was compared experimentally with the differential evolution on several test functions. The results showed that the proposed algorithm was more reliable and more effective in most of the test functions. Moreover, the proposed algorithm does not need a special tuning of its input parameters for each task.