Particle Swarm Optimization with Power-Law Parameter Based on the Cross-Border Reset Mechanism

Abstract

In order to improve the performance of traditional particle swarm optimization, this paper introduces the principle of Levy flight and cross-border reset mechanism. In the proposed particle swarm optimization, the dynamic variation of parameters meets the power-law distribution and the pattern of particles transition conforms to the Levy flight in the process of algorithm optimization. It means the particles make long distance movements in the search space with a small probability and make short distance movements with a large probability. Therefore, the particles can jump out of local optimum more easily and coordinate the global search and local search of particle swarm optimization. This paper also designs the cross-border reset mechanism to make particles regain optimization ability when stranding on the border of search space after a long distance movement. The simulation results demonstrate the proposed algorithms are easier to jump out of local optimum and have higher accuracy when compared with the existing similar algorithms based on benchmark test functions and handwriting character recognition system.

Keywords

• evolutionary computation
• optimization
• particle swarm optimization
• performance evaluation
• benchmark testing