An Fruit Fly Optimization Algorithm with Dimension by Dimension Improvement
To overcome the shortages of interference phenomena among dimensions, slow convergence rate and low accuracy, a new fruit fly optimization algorithm with dimension by dimension improvement is proposed. In addition, in order to speed up the algorithm conve
- PDF / 622,161 Bytes
- 12 Pages / 439.37 x 666.142 pts Page_size
- 59 Downloads / 265 Views
Abstract. To overcome the shortages of interference phenomena among dimensions, slow convergence rate and low accuracy, a new fruit fly optimization algorithm with dimension by dimension improvement is proposed. In addition, in order to speed up the algorithm convergence rate and avoid algorithm falling into local optimums, a Lévy flight mechanism is introduced to speed up the algorithm convergence rate and enhance the ability to jump out of the local optimum. The simulation experiments show that the proposed algorithm greatly speeds up the convergence rate and significantly improves the qualities of the solutions. Meanwhile, the results also reveal that the proposed algorithm is competitive for continuous function optimization compared with the basic fruit fly optimization algorithm and other algorithms. Keywords: Fruit fly optimization Multi-dimension function optimization Dimension by dimension improvement
1 Introduction Fruit fly optimization algorithm (FOA) [1, 2] was firstly proposed by Professor Pan in June 2011 and is a new global optimization algorithm inspired by the simulation of the foraging behavior of fruit flies. The fruit flies, by virtue of their better ability to smell and sight, feel various smell in the air to find food, and fly along the direction to food. FOA has been successfully applied to solve the function extreme value, the fine tuning Z-SCORE model coefficient, the generalized regression neural network parameter optimization and the support vector machine parameter optimization. Compared with other swarm intelligence algorithms, FOA has the advantages of simple algorithm, easy implementation of program code and less running time. Many improved versions of the FOA algorithm have been proposed. The application of combining the reverse learning strategy to the basic FOA algorithm [3], to a certain extent, improved the convergence rate and accuracy of the algorithm. Cheng and Liu proposed a new FOA algorithm based on chaotic map [4], the diversity of the population was improved by using chaos mapping. All mutated versions of FOA in iterative producing new solutions have one thing in common, that is, for a fruit fly, the solution is evaluated after the completion of the update for all dimensions. For multi-dimensional optimization problems, however, there exists a mutual interference phenomenon, the quality of solutions and the convergence rate of the algorithm will be © Springer International Publishing Switzerland 2016 D.-S. Huang et al. (Eds.): ICIC 2016, Part I, LNCS 9771, pp. 679–690, 2016. DOI: 10.1007/978-3-319-42291-6_68
680
H. Li et al.
affected using the evaluation strategy after the update to the values of all dimensions. Therefore, Zhong et al. proposed an iterative improvement strategy which can effectively avoid the mutual interference between dimensions [5]. In addition, dimension by dimension improvement strategy been used in some swarm intelligence algorithms, such as flower pollination algorithm [6] and cuckoo search algorithm [7]. To reduce the dependence of the algori
Data Loading...
