Intelligent Multiple Search Strategy Cuckoo Algorithm for Numerical and Engineering Optimization Problems

Abstract

This paper presents intelligent multiple search strategy algorithm (IMSS) as a new modification of cuckoo search (CS) to improve performance of the conventional algorithm. To do so, the proposed IMSS algorithm adopts a multiple search strategy and Q-learning technique. The introduced multiple search strategy couples CS and covariance matrix adaptation evolution strategy (CMAES) to explore search space more efficiently and also to reduce computational time of finding the optimal solution. More precisely, CS enables the IMSS to achieve better accuracy of final solutions through Levy flights, and CMAES enhances its convergence rate via a concept known as evolution path. To provide an intelligent balance between the exploration and exploitation behaviors, the IMSS employs Q-learning method and thereby acquires information about the performance of each search strategy. Then, it uses this information to dynamically select the best strategy for evolving candidate solutions as optimization process progress. In other words, the IMSS algorithm transforms the task of learning the optimal policy in Q-learning into the search for an efficient and adaptive optimization behavior. The IMSS is evaluated on CEC 2005 and CEC 2013 test suites, and its results are compared with results produced by several state-of-the-art algorithms. For further validation, the presented approach is also applied on two well-studied engineering design problems. The obtained results indicate that the IMSS provides very competitive results compared to other algorithms on the aforementioned optimization problems.