An Improved Shuffled Frog Leaping Algorithm with Comprehensive Learning for Continuous Optimization

摘要

This paper presents a shuffled frog leaping algorithm (SFLA) with comprehensive learning strategy (SFLA-CL) for global optimization. This algorithm uses a novel learning strategy whereby all other frogs' information of the memplex is used to update the worst frog's position. The strategy enables the diversity of the memplex to be preserved to discourage premature convergence. SFLA-CL also introduces a new search learning coefficient into the formulation of the original SFLA to enhance the convergence performance of SFLA. SFLA-CL has been evaluated, in comparison with existing evolutionary algorithm, such as SFLA, particle swarm optimization (PSO) and fast evolutionary programming (FEP), on five mathematical benchmark functions. Experimental results demonstrate that the SFLA-CL performs much better than SFLA, PSO, and FEP in optimizing these benchmark functions, particularly, in terms of its convergence rates and robustness.