A Study on Search Algorithms for Constructing Optimal Designs

บทคัดย่อ

Abstract

Computer simulated experiments (CSE) are often used in science and engineering applications. The nature of CSEis that they are time consuming and computationally expensive to run. Normally, the output response from computer simulated experiments is deterministic. Consequently the space filling designs, which focus on spreading design point sover a design space, are necessary. Latin hypercube designs (LHD) are normally practiced in the context of CSE. The optimal LHD for a given dimension of problem is constructed by using a search algorithm under a pre-specified optimality criterion. Usually this searching process takes a long time to terminate, especially when the dimension of the problem is large. This paper proposes methods to enhance the performance of search algorithms which are widely used in the context of CSE. The comparative studies are employed based on a range of problems and optimality criteria. The results indicate that the proposed method can improve the capability of the search algorithms for constructing the optimal LHD.

Keywords: Computer simulated experiments, Latin hypercube designs, simulated annealing algorithms, enhancedstochastic evolutionary algorithm, optimality criteria