Almost surely convergent global optimziation algorithm using noise-corrupted observations (Q1579659)

This papers deals with the random search approach for global optimization when the function is observed with noise. The authors propose a new hybrid algorithm based on random search and a local search method with one-sided randomized differences. The resulting algorithm is proved to converge almost surely to the global minimizer of the objective function. The authors present arguments which indicate that, in comparison with existing methods, the proposed algorithm not only requires much weaker conditions, but also is more efficient as shown by simulation. The results are compared with the article by \textit{S. Yakowitz} [SIAM J. Control Optimization 31, 30-40 (1993; Zbl 0769.62060)].