Modified Cuckoo Search Algorithm for Fast Convergence

Abstract

In recent years, several meta-heuristic optimization techniques have been generated. Recently introduced Cuckoo Search Algorithm (CSA), has proven its outstanding capabilities for solving optimization problems, such as increased convergence rate and greater global minimum values. An optimization algorithm termed as CSA is inspired by the lifespan of a Cuckoo bird. Unique lifestyle of this bird and their typical features of laying eggs and breeding have drawn inspiration for developing a new evolutionary optimization algorithm. Alike different evolutionary approaches, CSA starts with an initial set of population (i.e. cuckoo with eggs). There are two types of cuckoo population in different societies: mature cuckoos and cuckoo eggs. The origin of Cuckoo Search Algorithm is established from the struggle to survive among cuckoos. A fraction of cuckoos or their eggs get destroyed during the struggle of survival. But the survived cuckoos make a society and settle into a better habitat and they start laying eggs and reproducing there. Probably the survival attempt of cuckoos converges to a state that there is only one cuckoo society that exits with same profit values. As a novel evolutionary estimation technique, CSA has drawn much attention and extensive applications, due to its easy implementation. As most population-based algorithms, CSA is good at analyzing the promising area of the search space, but not so well at tuning the approximation to the minimization. In order to increase the efficiency and convergence rate of standard CSA, a modified cuckoo search algorithm has been introduced. Generally, all the parameters in cuckoo search algorithm are kept constant which results in the decreased efficiency of algorithm. To deal with this problem, the parameters of cuckoo search algorithm have been tuned by applying some tuning strategies to it. Considering various commonly used benchmark functions, numerical studies acknowledge that the modified algorithm can find better solution comparative to the solutions obtained by the standard algorithm. On account of this, it is expected that the modified algorithm can be applied successfully to a broad variety of optimization problems.