应用随机游走和Boltzmann的烟花爆炸优化算法

Improved Fireworks Explosion Optimizer Using Randomly Walk and Boltzmann

摘要: 为了克服烟花爆炸优化(Fireworks Explosion Optimization, FEO)算法容易早熟、解精度低的弱点，提高算法的收敛速度，提出了一种应用Lévy Flight机制的改进烟花爆炸优化(LFBFEO)算法。改进算法应用佳点集进行种群初始化，并在爆炸变异算子中引入Lévy Flight机制产生新个体，引导个体加速向最优个体靠近，同时使个体具有摆脱局部极值约束的能力。应用Boltzmann机制实施子个体的生存选择，以一定的概率接受劣质个体，使算法避免早熟。最后，在6组标准测试函数上的实验表明，LFBFEO克服了FEO算法的不足，一定程度上提高了算法的收敛速度和收敛精度。

Abstract: To overcome the prematurity and low precision of Fireworks Explosion Optimizer (FEO) and improve its convergence speed, an improved FEO (LFBFEO) using Lévy Flight and Boltzmann mechanism was proposed. LFBFEO initializes the population by the good-point technique and uses Lévy Flight to improve the explosion operator of FEO, which not only can guide the individuals’ approach optimization fleetly, but also make the individuals have the ability to avoid the region of local optimization. Moreover, we introduce the Boltzmann mechanism into the selection operator. This mechanism makes some not-so-bad individuals accepted and makes the algorithm avoid premature. Finally, experiments were conducted on a set of 6 classical test functions. The experimental result demonstrates that the LFBFEO improves the FEO and to some extent it improves the algorithm convergence speed and accuracy.