Adaptive Cuckoo Search Based on Evaluation and Control of Search State

Abstract

In this paper, to improve the search performance and usability of Cuckoo Search (CS), we develop a new Adaptive Cuckoo Search (ACS) based on evaluation and control of its search state by adjusting one of its parameters. ACS consists of a new evaluation indicator and a feedback control mechanism. Using the results of the quantitative analysis of the relationship between the parameter and random numbers in Lévy Flight, we clarify the effects of the parameter on the search dynamics of CS in terms of the metaheuristics strategy (diversification / intensification). After the analysis, the indicator which evaluates the search state is defined. The mechanism makes an ideal state (a target of the indicator) and controls the search state to the ideal state by adjusting the parameter. Overall, ACS is expected to improve the search performance of CS by adequately realizing diversification and intensification during a search. The performance and adaptability of ACS are verified through numerical simulations for 9 typical benchmark functions.