A Swarm Intelligence Based (SIB) method for optimization in designs of experiments

Natural heuristic methods, like the particle swarm optimization and many others, enjoy fast convergence towards optimal solution via inter-particle communications. Many applications of such methods are applied to the optimization in engineering, but only a few to the optimization in statistics. It is especially difficult to implement in the optimization problems of experimental designs as the search space is mostly discrete, while most natural heuristic methods are limited to searching continuous domains. This paper introduces a new natural heuristic method called Swarm Intelligence Based method for optimizing problem with a discrete domain. It includes two new operations, MIX and MOVE, for combining two particles and selecting the best particle respectively. This method is ready for the search of both continuous and discrete domains, and its global best particle is guaranteed to monotonically move towards the optimum. Several demonstrations on the optimization of experimental designs are given at the end of this paper.