Research On Electromagnetism-like Mechanism Algorithm

The study of global optimization has got more and more people’s attention both insocial life and scientific research field in recent years, thus it has become an importantresearch focus now. However, it is difficult to solve the global optimization with thetraditional deterministic algorithms for the mathematical models constructed by themeither have high dimensions or don’t have good analytic properties. Therefore, someintelligent optimization algorithms have emerged, such as genetic algorithm, ant colonyalgorithm, particle swarm optimization, simulated annealing algorithm,electromagnetism-like mechanism algorithm and etc. These algorithms always have lessneed even no need the analytic properties of the objective functions when searching forthe global optimal values. But it is still necessary to design a robust and effectivealgorithm for solving the high-dimensional and complex functions.Aim to solve the unconstrained optimization problems, a meta-heuristic algorithmnamed as electromagnetism-like mechanism algorithm is studied in this paper. And then,an improved algorithm which is Electromagnetism-like Mechanism algorithm based ongood point set is proposed. The improved algorithm constructs the initial populationusing the concepts of good point set in number theory to make the initial particles moreevenly distributed in the feasible region. It uses the chaos search method instead ofrandom linear search algorithm, which can not only avoid the premature convergence,but also improve the accuracy of the solution. The algorithm uses an adaptive moveoperator to update the location of the charged particles and adds a genetic coefficient tothe update formula, so that the particles have more chance to move into the otherfeasible regions.Experimental results show that the improved algorithm can converge to globaloptimal values more effectively and accurately, and meanwhile can solve the generalproblems with bounded variables well, as well as nondifferential, high dimensional,multimodal and even singular problems.In further research, the Electromagnetism-like Mechanism algorithm would beimproved or combined with other algorithms to make it converge much faster, obtainmore accurate solutions. Meanwhile the improved algorithm will be suited to solve theoptimization problems in actual project.