A distributed search algorithm for global optimization on numerical spaces

This article presents a new algorithm that searches for the global extrema of numerical functions of numerical variables. This ‘Distributed Search’ algorithm builds an evolving ‘visiting’ probability distribution on the search domain, and the process converges towards states in which the probability density is maximal on the neighborhood of the target extrema. The convergence of the algorithm is demonstrated. Then its performance is tested on some ‘hard’ test functions and compared to that of a recent, well-known algorithm.