The Research Of Electromagnetism-like Mechanism And Its Application

Global optimization is very widely used in many fields, many optimization problems in engineering can be treated as global optimization problems, and this greatly leads the advance and development of global optimization techniques. In the field of global optimization, a typical population-based stochastic search method got great attention and development.In this paper, we study a new population-based global optimization method—Electromagnetism-like Mechanism (EM). The method simulates the attraction-repulsion mechanism in electromagnetism to move sample points toward optimality and is thus referred to as electromagnetism-like mechanism algorithm. The major task of this paper is researching EM, modifying it and applying it in some fields, such as function optimization with high dimensionality, neural network training and flow shop scheduling.Firstly, the theory and the scheme of EM algorithm are generally presented. Secondly, In the field of function optimization, on the base of the traits and mechanism of EM, the modified EM is proposed—the equation of total force is modified and a moving coefficient is introduced. Through testing the benchmark and comparing with other classical methods, we reach the conclusion that EM is robust and has good perspective.Then, in order to overcoming the bad traits of some traditional training algorithms of neural network such as very slow convergence rate and the nature of converging locally, the training algorithm of neural network based on EM is presented and the trained neural network is successfully applied in Iris, Ionosphere and Breast cancer classification problem. Moreover, through replacing several worse particles we apply EM to flow shop scheduling problems by random key representation, then it links EM with discrete optimization problems.Finally, the conclusion and perspective of EM is given.