A UNIVERSAL MULTIMETHOD SEARCH STRATEGY FOR COMPUTATIONALLY EFFICIENT GLOBAL OPTIMIZATION

In the last few decades many different algorithms have been developed for solving complex search and optimization problems. The focus has been on the development of a single universal genetic operator for population evolution that is always efficient for a diverse set of optimization problems. However, existing theory and numerical experiments have demonstrated that it highly unlikely that such a universal operator exists. In this paper we show that significant advances in the field of evolutionary computation can be made if we embrace a concept of self-adaptive multimethod optimization, the goal of which is to develop a combination of search methods that have all the desirable properties to efficiently handle a wide variety of response surfaces. We present a new optimization algorithm, called AMALGAM-SO, that implements this new concept of multimethod search, and implements a restart strategy with successively increasing population size. Benchmark results in various dimensions, including real-world applications show that AMALGAM-SO is generally superior in efficiency, robustness and reliability to currently available search algorithms. The new search method is relatively easy to implement, and is designed to take full advantage of the power of distributed computer networks.