Research On Shuffled Frog Leaping Algorithm For Continuous Optimization Problem

In recent years, with the flying development of productivity and science technology,a large quantity of optimization problems existing in practical engineering applicationhave been becoming increasingly complex. However, it is difficult for traditionaldeterministic algorithms to solve the complicated global optimization problems whichare non-convex and non-differentiable. Since these heuristic algorithms have greatintelligence and high efficiency, a lot of scholars focus on the research and developmentof them. As a new meta-heuristic, Shuffled Frog Leaping Algorithm （SFLA）is apopulation-based cooperative search metaphor inspired by natural memetics, and hasbeen developed for solving global optimization problems. SFLA has many advantages,such as simple structure, high convergence rate and strong global search capability. Butfor some complex problems in practical engineering, SFLA still has the disadvantagesof large probability of falling into local optimum and low convergence rate in laterevolution process.In the paper, an Improved Shuffled Frog Leaping Algorithm（ISFLA）is proposedto eliminate the drawbacks of the original SFLA. The initial population of SFLA isgenerated by using the random uniform design method, therefor, the initial frogs can bemore even and the algorithm’s probability of searching to the global optimal value canbe increased. During the different phase of the running process of ISFLA, a dynamicinfluence factor is introduced to adjust the local search strategy adaptively, which canmake it be more in line with the dynamic characteristics of the population search andmore efficient. Besides, the variance of the population’s fitness is calculated to judgewhether the population falls into local optimum, and once the population falls into localoptimum, the Dual Chaos Optimization Mechanism will be applied in local search toimprove the diversity of the population, which will deal with the premature convergenceand improve the accuracy of the solution.Some benchmark functions are simulated with ISFLA, and the results show thatISFLA has better optimizing performance. The improved algorithm can not onlyimprove the convergence rate, but also deal with premature convergence and obtainbetter accuracy of solutions.In addition, SFLA in itself should be improved in the further research. For theimportance of the settings of parameters for the performance of SFLA, further research will focus on the settings and relationships among parameters and the application of theimproved algorithm in solving the optimization problems in practical engineering.