Shuffled Frog Leaping Algorithm with Adaptive Exploration

Shuffled frog leaping algorithm is a nature inspired memetic stochastic search method which is gaining the focus of researchers since it was introduced. SFLA has the limitation that its convergence speed decreases towards the later stage of execution and it also tends to stuck into local extremes. To overcome such limitations, this paper first proposes a variant in which a few new random frogs are generated and the worst performing frogs population are replaced by them. Experimental results show that a high number of replaced frogs does not always provide better results. As the execution progresses the optimized number of replaced frogs decreases. Based on the experimental observations, the paper then proposes another variant in which the number of replaced frogs adapts to the stage of the execution and hence provides the best results regardless of the stage of execution. Experiments are carried out on five benchmark test functions.