Comparison of Three Well-known Global Optimization Algorithms for Solving Black Box Problems

Abstract

Stochastic Global Optimization Algorithms (GOA) have been employed widely due to their efficiency in handling difficult engineering optimization problems. The stochastic or heuristic methods are based on the random generation of feasible points, or sampled points, and non-linear local optimization search procedures using these points. In addition to a number of well known global optimization methods, many new GOA have been presented recently for numerous optimal design applications. In this paper, three well-known global optimization algorithms include Simulated Annealing (SA), Differential evolution (DE), and Harmony search (HS) are closely tested and compared. The historical development, special features and trends on the development of these stochastic GOA are studied. Special consideration is devoted to the performance of those GOA while the number of iterations are limited. Commonly used benchmark optimization functions are used as test examples to reveal the pros and cons of those three recognized global optimization algorithms.