Abstract:

To overcome the problem of premature and low precision in Teaching and Learning Based Optimization（TLBO） algorithm, this paper proposes an improved TLBO with Dynamic Self-adapting Learning（DSLTLBO）. In the “teaching” phase of DSLTLBO, a factor based on the self-adaptive is introduced, which makes the current individual learn mainly from the best individual in the early stage and Maintain one’s own state in the later for keeping the diversity of population. Later stage in iteration, teacher individual excutes the Dynamic Random Searching（DRS）to improve its ablity of exploring the new solutions around them. Experiments on 10 classical Benchmark functions show that DSLTLBO  has better convergence speed and solution accuracy than TLBO, and is suitable for solving higher dimensional optimization problems.

Key words: Teaching and Learning Based Optimization（TLBO）, dynamic self-adaptive, learning factor, Dynamic Random Searching（DRS）

摘要：

为了克服教与学优化（TLBO）算法容易出现早熟和解精度低的问题，提出了一种动态自适应学习的改进教与学优化（DSLTLBO）算法。在DSLTLBO算法的“教”阶段，引入一个自适应变化的因子，使当前个体在早期主要向最优个体学习，后期能够较好地维持自身状态，种群多样性得以保持。在算法的后期，教师个体通过执行动态随机搜索算法，提高最优个体勘探新解的能力。在10个经典的Benchmark函数上的实验表明，该算法具有较好的收敛速度和解精度，较标准TLBO有较大能力提升，适合于求解较高维度的优化问题。

关键词: 教与学优化（TLBO）, 动态自适应, 学习因子, 动态随机搜索（DRS）