The project aims the research of benchmarking techniques for derivative-free optimization methods. The two main directions in the development of these optimization methods are mathematical programming (e.g. the DIRECT method) and evolutionary algorithms (e.g. the differential evolution algorithm). The term benchmarking refers to a set of procedures for comparing such methods. Appropriately chosen benchmarking techniques can reveal the structural bias of some methods, or create guidelines for choosing suitable methods for a given optimization problem. Some of the current issues in this field are the strong emphasis on artificially created benchmark sets, the structural problems of some sets and algorithms, and the small intersection between benchmarking techniques and the comparison of methods from the two main development directions mentioned above.

Klíčová slova
global optimization;derivative-free methods;benchmarking