An Adaptive Shuffled Frog Leaping Algorithm

摘要

For the defects of Classic Shuffled Frog Leap Algorithm such as premature, slow convergence rate and low precision, this paper studies an adaptive SFLA (ASFLA) with the amelioration on population initialization, subpopulation division, and partial navigation. Population initialization is improved by opposite strategy based on disturbance factor which enriches the diversity of population. Subpopulation division is improved by fore and aft synchronization, narrowing the differences among subpopulations. Partial navigation is improved by the design of combination of adaptive inertia factors and gradient information, accelerating the convergent rate. Experimental results show that the ASFLA is superior to SFLA. The ASFLA which can obtain a higher globally optimal solution accuracy by a higher convergent rate avoids the premature convergent phenomenon.