On optimization and extreme value theory (Q1405352)

A stochastic optimiser is an algorithm ruled by probabilistic procedures which aims at finding the optimum point of an objective function. The authors study the fit of the optimization outcomes to an extreme value distribution and suggest that the parameters of that distribution characterize the performance of the optimiser. Simulation experiments using random search and two evolution strategies are presented. The empirical results for the random search strategy are supported by theoretical results.