The Algorithm of Statistical Global Optimization with Search Step Adaptation

摘要

We consider the algorithm of the statistical global optimization based upon the Rastrigin cone method, and also the general strategy of search of the global minimum of multiextremal functions and functions with ravines. Particularly, we attend to the choice problem of optimization parameters. We propose the criteria and procedure of the search step adaptation for an objective function with narrow ravines. As an example, we have considered the test tasks.