The Electromagnetic Mechanism Algorithm. Class Research And Improvements

Electromagnetism-like Mechanism Algorithm is a population-based heuristic search algorithm, based on the attracting-repelling characteristics of the electrical charges. Driven by the total force, all particles move in a random step. As the forces between particles are calculated, the distance between particles has too much impact on the forces between them, resulting in EM algorithm's own weakly local search capability. The introduction of the local search increases the time complexity of the algorithm.In this paper, based on the research trends of EM algorithm, EM's defects are analyzed from three respects, i.e., population, force calculation formula and moving patterns. In addition, the impact of local search on algorithm performances is also explained. After the analysis and explanation, two improved EM algorithm are introduced, as well as several ideas about population are given. The main research works are as follows: (1) a population-updating EM algorithm, EM-WPU, is introduced, based on elimination mechanism and disturbance factor. In EM-WPU, after each iteration, population is updated, generating a new population better to search. (2) An improved EM algorithm, EM-WLS, is presented. In EM-WLS, different force formulas are constructed according to the size of distances between particles, so as to weaken the impact of distance on the forces. At the same time, through the introduction of weighting parameter of component of forces, effects of attractive/ repulsive function on the total force are adjusted. The direction of particle movement by total force or by deviating from the force direction is determined by the angle between the total force on particles and the optimal particle's attraction on the particle. In addition, the improved EM algorithm also removes the local search part of the EM algorithm. (3) This paper comprehensively describes the characteristics of biological populations in nature, and further analyzes the defects of algorithm population, providing some improvement thoughts. Experimental results show that the two improved EM algorithm improves the accuracy of solutions and can better solve the high-dimensional optimization problems.