An Improved Dolphin Swarm Algorithm Based on Kernel Fuzzy C-Means in the Application of Solving the Optimal Problems of Large-Scale Function

Weibiao Qiao,Zhe Yang

doi:10.1109/access.2019.2958456

Abstract

The solution of high dimensional function has always been a hot topic. In this paper, a novel algorithm based on Kernel Fuzzy C-means and dolphin swarm algorithm are proposed to solve high-dimensional functions more accurately. First, to improve the global convergence ability of dolphin swarm algorithm, Kernel Fuzzy C-means is introduced into the algorithm, named as Kernel Fuzzy C-means dolphin swarm algorithm (KFCDSA); Second, the five typical high-dimensional functions are applied to test the performance of the combination of KFCDSA. Finally, some indicators are used to evaluate the performance of different meta-heuristic algorithms. The results show that: the performance of the proposed algorithm exceeds that of the dolphin swarm algorithm and some advanced metaheuristic algorithms considered for comparison based on five different evaluating indicators. Through the test results, it can be concluded that introducing Kernel Fuzzy C-means into dolphin swarm algorithm is an effective improvement and provides a possibility for obtaining global optimal solutions for high-dimensional functions.

Highlights

The swarm intelligence algorithm is a new evolutionary computing technology [1]–[6]
The results show that Kernel Fuzzy C-means dolphin swarm algorithm (KFCDSA) has better convergence ability than other algorithms considered for comparison
KFCDSA is 186.5158, 500.1273, 2328.4746, 2938.5560, 2503.4878, and 723.7068 lower than dolphin swarm algorithm (DSA), WOA, AGA, APSO, WSA, CSA in view of min. These results suggest that KFCDSA has better performance than DSA, WOA, AGA, APSO, WSA and CSA to solve Levy function

Summary

Introduction

The swarm intelligence algorithm is a new evolutionary computing technology [1]–[6]. Many optimization problems often have high dimensions, which are called large-scale global optimization problems [8]. Been many meta-heuristic algorithms, which are proposed to improve the accuracy and efficiency of large-scale global optimization. With the gradual increase of largescale global optimization decision variables, the search space increases exponentially with the number of decision variables, and meta-heuristic algorithms usually require too much computational cost, especially the optimization of highdimensional functions. Based on existing metaheuristic algorithms, it is of great significance to develop an improved meta-heuristic algorithm to solve high-dimensional function optimization problems

Objectives

Results

Conclusion