College of Computer Science, Chongqing University, Chongqing 400044, China.
Key Laboratory of Dependable Service Computing in Cyber Physical Society, Ministry of Education, Chongqing 400030, China.
Math Biosci Eng. 2022 Mar 23;19(5):5241-5268. doi: 10.3934/mbe.2022246.
In the traditional particle swarm optimization algorithm, the particles always choose to learn from the well-behaved particles in the population during the population iteration. Nevertheless, according to the principles of particle swarm optimization, we know that the motion of each particle has an impact on other individuals, and even poorly behaved particles can provide valuable information. Based on this consideration, we propose Lévy flight-based inverse adaptive comprehensive learning particle swarm optimization, called LFIACL-PSO. In the LFIACL-PSO algorithm, First, when the particle is trapped in the local optimum and cannot jump out, inverse learning is used, and the learning step size is obtained through the Lévy flight. Second, to increase the diversity of the algorithm and prevent it from prematurely converging, a comprehensive learning strategy and Ring-type topology are used as part of the learning paradigm. In addition, use the adaptive update to update the acceleration coefficients for each learning paradigm. Finally, the comprehensive performance of LFIACL-PSO is measured using 16 benchmark functions and a real engineering application problem and compared with seven other classical particle swarm optimization algorithms. Experimental comparison results show that the comprehensive performance of the LFIACL-PSO outperforms comparative PSO variants.
在传统的粒子群优化算法中，粒子在种群迭代过程中总是选择从表现良好的粒子中学习。然而，根据粒子群优化的原理，我们知道每个粒子的运动都会对其他个体产生影响，即使是表现不佳的粒子也可以提供有价值的信息。基于此考虑，我们提出了基于 Lévy 飞行的逆自适应综合学习粒子群优化算法，简称 LFIACL-PSO。在 LFIACL-PSO 算法中，首先，当粒子陷入局部最优而无法跳出时，采用逆学习，并通过 Lévy 飞行获得学习步长。其次，为了增加算法的多样性并防止其过早收敛，采用综合学习策略和环形拓扑作为学习范例的一部分。此外，使用自适应更新来更新每个学习范例的加速度系数。最后，使用 16 个基准函数和一个实际的工程应用问题来衡量 LFIACL-PSO 的综合性能，并与其他七个经典的粒子群优化算法进行比较。实验比较结果表明，LFIACL-PSO 的综合性能优于比较的 PSO 变体。
