Research And Application Of Niche Shuffled Frog Leaping Algorithm

Owing to the complexity of practical engineering problems, it is very difficult tosolve a large number of optimization problems. In recent years, intelligent optimizationalgorithms based on intelligent behavior of biological communities have attracted moreand more attention of researchers, because these algorithms do not rely on the gradientinformation of problems and have the potential to jump out the local optimum point.Shuffled leap frog algorithm (SFLA) is an intelligent optimization algorithm whichsimulates the feeding behavior of frogs. This algorithm with a few parameters is robust,simple and easy to understand. But for solving some complex problems, SFLA is stillslow in convergence rate and easy to fall into a local optimal value.For the defects of classic shuffled leap frog algorithm such as premature, slowconvergence rate and low precision, this paper studies the optimization mechanism ofSFLA and proposes a shuffled frog leaping algorithm using niche technology namedniche shuffled frog leaping algorithm (NSFLA). The new algorithm applies theRestricted Competition Selection (RCS) niche technology to make each sub-populationdynamically form search spaces which are independent of one another, restraining theconvergence because of group collaboration, and making the algorithm with betterglobal optimization ability and convergence speed. In the updating formula for thesolution, an adaptive factor which is designed to adjust the moving step can guide thesolution toward the optimal solution, which speeds up the convergence rate. Furthermore,the population elimination mechanism is used in the algorithm to randomly initialize thesub-population which has fallen into the local optimum, which avoids the prematureconvergence of the algorithm.Experimental results show that the NSFLA in this paper is superior to SFLA.NSFLA which can avoid the premature convergence effectively improves theoptimization accuracy and convergence speed.How to select and adjust the control parameters is the key to affect the performanceof the NSFLA. In subsequent studies, setting parameter in the NSFLA will beresearched, in order to continue to improve the performance of NSFLA. In addition, thedesign methods of initial population and local search strategies still can be improved.