Optimisation based on 'Nearest Better Neighbor': A Preliminary study

11th Curtin University Technology, Science and Engineering (CUTSE) International Conference

工业技术（总论）

Wong, W.K.^1

Department of Electrical and Computer Engineering, Faculty of Engineering and Science, Curtin University Malaysia, Miri Sarawak

98009, Malaysia^1

关键词: Benchmark functions;    Best population;    Early convergence;    Iteration periods;    Optimisations;    Random mutation;    Research reports;    Solution space;   

Others  :  https://iopscience.iop.org/article/10.1088/1757-899X/495/1/012018/pdf
DOI  :  10.1088/1757-899X/495/1/012018

学科分类：工业工程学

来源: IOP

【 摘 要 】

A multi agent optimization was proposed and evaluated in this research report. It incorporates multi agent search by learning from the 'nearest better solution'. Each agent learns from the nearest agent with better solution thereby ensures that the optimization explores solution spaces between the solutions. Weightage multipliers determine how much each solution segment adapts to the 'nearest better solution' based the ratio of the 'nearest better solution fitness' to the best population fitness. A small amount of random mutation is introduced in every iteration to prevent early convergence. A converging time period (50% of last iteration period) is assigned in order for the agents to converge to a specified range in which the random mutation mechanism is totally repressed. Results showed that the algorithm is able to produce comparable results as compared with other algorithm such as Genetic Algorithm (GA) and particle swarm optimization (PSO). Evaluations were performed using 11 benchmark functions based on average ranking.