融合涡流搜索和差分策略的教学优化算法  

Teaching-learning-based Optimization Integrating Vortex Search and Differential Strategy

作　　者：李子扬 刘宗堡[1] LI Zi-yang;LIU Zong-bao(School of Earth Sciences,Northeast Petroleum University,Daqing 163318,China)

机构地区：[1]东北石油大学地球科学学院,黑龙江大庆163318

出　　处：《计算机技术与发展》2022年第2期183-189,共7页Computer Technology and Development

基　　金：黑龙江省优秀青年科学基金(YQ2020D001);中国石油科技创新基金(2020D-5007-0102)。

摘　　要：模拟课堂教学行为的教学优化算法,具有操作简单且搜索能力强的突出优点。尽管该算法为增强种群多样性采取了消除重复个体操作,然而在算法后期依然容易陷入早熟收敛。为提高教学优化算法的搜索能力,该文通过融合涡流搜索和差分进化这两种策略,提出了改进措施。改进后的算法包括:教师自学、向教师学、学生互学三种行为。首先,在每轮循环的开始,增加了基于涡流搜索的教师自学习行为,从而使作为最优个体的教师也获得一定的改进机会。其次,在教师阶段和学生阶段的个体更新式中,均增加了体现不同个体之间差异的差分算子,同时在学生阶段增加了轮盘赌选择策略,以便使优良个体获得更多更新的机会。10个标准测试函数的仿真结果表明,改进算法的寻优能力不仅比原算法有大幅度提升,而且对于部分复杂函数也优于同类其他对比算法,从而揭示出通过融合涡流搜索和差分策略提升教学优化算法性能的研究方案是可行的。The teaching-learning-based optimization algorithm that simulates classroom teaching behavior has the outstanding advantages of simple operation and strong search ability. Although the algorithm takes the operation of eliminating duplicate individuals to enhance the diversity of the population, it is still easy to fall into premature convergence in the later stage. In order to improve the search ability of the teaching-learning-based optimization, we propose improvement measures by integrating the two strategies of vortex search and differential evolution. The improved algorithm includes three behaviors: self-learning by teachers, learning from teachers, and mutual learning by students. First of all, at the beginning of each cycle, the teacher’s self-learning behavior based on vortex search is added, so that the teacher who is the best individual can also get a certain chance of improvement. Then, in the individual update formulas of teacher stage and student stage, the differential strategies reflecting the differences between different individuals are introduced. At the same time, the roulette selection strategy is employed in the student stage, so that the excellent individuals can get more opportunities to update. The simulation results of 10 benchmark functions show that the optimization accuracy of the proposed algorithm is not only greatly improved compared with the original algorithm, but also outperforms other comparable algorithms of the same kind for some complex functions. It is showed that it is feasible to improve the performance of teaching-learning-based optimization by integrating vortex search and differential strategies.

关 键 词：教学优化 涡流搜索 差分策略 轮盘赌选择 算法设计 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]