Abstract:

An information interactive Particle Swarm Optimization （PSO） algorithm for teaching and learning was proposed to solve high dimensional problems of low convergence rate and lack of diversity in a single population. The population was divided into two subpopulations dynamically according to evolutionary process， and processed by PSO algorithm and teaching and learning based optimization algorithm respectively. At the same time， learner stage was used by the particles to carry out information interaction between subpopulations， and by evaluating convergence and diversity indexes， the convergence ability and diversity of particles were balanced in evolutionary process. Compared with PSO algorithm， hybrid PSO and Grey Wolf Optimizer （GWO） algorithm， and improved GWO algorithm using nonlinear convergence factor and elite re-election strategy and other evolutionary algorithms in different dimensions of 15 standard test functions， the proposed algorithm can converge to the theoretical optimal value on multiple test functions， which is 1 to 6 times faster than other algorithms. Experimental results show that the proposed algorithm has good convergence accuracy and speed.

Key words: Particle Swarm Optimization (PSO) algorithm, teaching and learning optimization algorithm, dynamic population adjustment, information interaction, normalization method, multi-population collaboration

摘要：

针对单一种群在解决高维问题中收敛速度较慢和多样性缺失的问题，提出了一种教与学信息交互粒子群优化（PSO）算法。根据进化过程将种群动态地划分为两个子种群，分别采用粒子群优化算法和教与学优化算法，同时粒子利用学习者阶段进行子种群之间信息交互，并通过评价收敛性和多样性指标让粒子的收敛能力和多样性在进化过程中得到平衡。与粒子群优化算法、混合灰狼粒子群算法、重选精英个体的非线性收敛灰狼优化（GWO）算法等多个进化算法在15个标准测试函数的不同维度下进行对比实验，所提算法在多个测试函数上可以收敛到理论最优值，速度相对于其他算法提高了1~6倍。实验结果表明，所提算法在收敛精度和收敛速度上具有较好的效果。

关键词: 粒子群优化算法, 教与学优化算法, 种群动态调整, 信息交互, 归一化方法, 多种群协同