| 摘 要: 教与学优化算法受教学现象的启发而提出，具有收敛速度快和模型参数少的优势。为了提高算法的收敛精度和全局搜索能力，提出预测选择教与学优化算法。首先，采用反向学习机制初始化种群个体位置，保持种群多样性。然后，在“教阶段”设置惯性权值和加速系数，提高算法的运算速度和解的质量。最后，在“学阶段”并行引入三种种群个体预测机制，选择最佳适应度值的个体进行下一次的寻优，提高算法全局搜索能力。通过基准数学函数测试得到的实验结果表明：改进算法的收敛精度和解的质量优于原始教与学优化算法。 |
| 关键词: 教与学优化算法；预测选择；反向学习机制；收敛精度 |
|
中图分类号: TP181
文献标识码: A
|
| 基金项目: 天津市自然科学基金(20JCQNJC00430)；大学生创新创业训练项目(202110069003，202110069034，202010069066). |
|
| Research and Application of Prediction-selection Teaching-learning-based Optimization Algorithm |
|
MA Yunpeng, LIU Shilin, DONG Wenju, XIE Siqi, WANG Ziyan, LV Dinglian
|
(School of Information Engineering, Tianjin University of Commerce, Tianjin 300134, China)
mayunpeng@tjcu.edu.cn; 1595996480@qq.com; 2458397196@qq.com; 2632614453@qq.com; mctldl@qq.com; 3202013353@qq.com
|
| Abstract: Teaching-learning-based Optimization (TLBO) algorithm with fast convergence speed and less parameter, is proposed based on teaching-learning practice. In order to improve convergence accuracy and global search ability of the algorithm, a Prediction-selection Teaching-learning-based Optimization (PSTLBO) algorithm is proposed. First of all, reverse learning mechanism is used to initialize the position of population individuals and to maintain population diversity. Then, inertia weight and accelerating factor are set in teaching phase to improve calculation speed and solution quality. Finally, in learning phase, three kinds of population prediction mechanisms are introduced in parallel and the individual with the best fitness value is selected for the next optimization, which improves the global search ability of the algorithm. Through benchmark mathematical function test, the experimental results show that convergence accuracy and solution quality the improved algorithm are better than the original teaching-learning optimization algorithm. |
| Keywords: teaching-learning-based optimization algorithm; predictive selection; reverse learning mechanism; convergence accuracy |
