Analysis of the Effect of Search Step Size on the Accuracy and Convergence Properties of Electromagnetism-Like Mechanism Algorithm

摘要

An Electromagnetism-like Mechanism algorithm (EM) is a meta heuristic search algorithm known for its ability to solve complex global optimization problems. In a standard EM, the size of the search steps is randomly determined. In this paper, an investigation is carried out to analyze the effects of search step size on the overall convergence performance of EM. A standard EM is modified into two different sets of algorithms with different extremes of search step sizes. The modified algorithms are then tested in 10 standard benchmark functions to evaluate the effect of search step sizes on the performance of each. The results show that the modified EM with smaller search steps performed a much more detailed search for possible solutions in all dimensions. This enabled the overall convergence to achieve relatively higher accuracy in objective values.