Island-based Cuckoo Search with Highly Disruptive Polynomial Mutation

Abstract

The island model is one of the most well-known structured population strategies used to control the diversity in evolutionary algorithms. The population of an island-based evolutionary algorithm is normally partitioned into several sub-populations (islands). An evolutionary
algorithm is then applied to each island independently. After a number of predefined generations, a migration process takes place to exchange specific candidate solutions between the islands. Recently, the Cuckoo search (CS) algorithm has been proposed as a population-based algorithm that mimics the nesting and parasitic reproduction behaviors of some cuckoo species. The main drawback of the CS algorithm is that its evolutionary operators may not adequately preserve the diversity of its population during the evolution process which may cause it to converge earlier than expected to suboptimal solutions. This paper introduces an improved variation of CS called island-based CS with polynomial mutation (iCSPM) that adapts two improvements to CS. First, the strategy of island model is incorporated into the CS algorithm to empower its capability in controlling the diversity of its population. Second, the L´evy flight method in CS is replaced with the highly disruptive polynomial mutation method in an attempt to enhance the exploration of CS. The iCSPM algorithm was evaluated using 15 standard benchmark functions in terms of the accuracy and reliability of the obtained results over multiple simulations. The sensitivity analysis of the main parameters of iCSPM was carried out to show their effect on the convergence behavior of iCSPM. The experimental results suggest that iCSPM provides more accurate and reliable results than 3 competitive methods. The source code of iCSPM is available at https://www.dropbox.com/sh/99x5374fiz2e390/AAC_6Eb9VFrDEdvt6wAmJstSa?dl=0