Abstract: Teaching-learning-based optimization (TLBO) is a recently proposed swarm intelligent algorithm that simulates the process of teaching and learning. Concerning the problems that TLBO is easy to premature, low solution precision, slow convergence speed of weakness, this paper proposes an improved TLBO named S-TLBO (small world neighborhood TLBO). S-TLBO adopts small world network as its spatial structure, and individuals of S-TLBO is looked as the nodes of network. In teaching phase, student individuals learn from teacher individual based on probability, and they learn from their neighbor nodes which are better in learning phase. The best in dividual executes opposition based learning (OBL) algorithm to exploiting and exploring. Some experiments are conducted on many classical testing functions, the results show that the improved algorithm has superior global convergence and higher precision, especially fits for solving multimode and high dimension function optimization problems.

Key words: teaching-learning-based optimization (TLBO), small world network, neighborhood structure, opposition based learning (OBL)

摘要： 教与学优化（teaching-learning-based optimization，TLBO）算法是近年来提出的一种通过模拟“教”与“学”行为的群体智能算法。为了克服教与学优化算法容易早熟，解精度较低，后期收敛速度慢等弱点，提出了一种改进的教与学优化算法，并命名为S-TLBO（small world neighborhood TLBO）。该算法采用小世界网络作为其种群的空间结构关系，种群中的个体被看作是网络上的节点。在算法的“教”阶段，学生基于概率向教师个体进行学习，而在“学”阶段，学生则在自己的邻居节点中随机选择较为优秀的个体进行学习。为了提高加强算法的勘探新解和开采能力，引入教师个体执行反向学习算法。在多个经典的测试函数上的实验结果表明，所提出的改进算法具有较高的全局收敛性和解精度，适合于求解较高维度的多模态函数优化问题。

关键词: 教与学优化（TLBO）, 小世界网络, 邻域结构, 反向学习（OBL）