混沌自适应非洲秃鹫优化算法训练多层感知器  

Chaotic adaptive African vulture optimization algorithm for training multilayer perceptron

作　　者：申晋祥[1] 鲍美英[1] 张景安 周建慧 SHEN Jin-xiang;BAO Mei-ying;ZHANG Jing-an;ZHOU Jian-hui(School of Computer and Network Engineering,Shanxi Datong University,Datong 037009,China;Network Information Center,Shanxi Datong University,Datong 037009,China)

机构地区：[1]山西大同大学计算机与网络工程学院,山西大同037009 [2]山西大同大学网络信息中心,山西大同037009

出　　处：《计算机工程与设计》2024年第2期546-552,共7页Computer Engineering and Design

基　　金：国家自然科学基金项目(11971277);山西大同大学科研基金项目(2020k10);山西大同大学云冈专项基金项目(2020YGZX016);山西大同大学校级教学改革创新基金项目(XJG2021249)。

摘　　要：针对训练多层感知器(MLP)时,算法对初始值敏感、易陷入局部最优和收敛速度慢等问题,对新型启发式算法非洲秃鹫优化算法提出改进算法IAVOA。在初始化种群时引入Logistic混沌映射,增加种群的多样性;对最优秃鹫和次优秃鹫增加自适应权重系数,自动调整这两类秃鹫对普通秃鹫的引导作用;IAVOA用于MLP的训练,采用均方误差的平均值作为适应度函数寻找MLP的连接权重和偏差的最佳组合。选取4个不同复杂度的分类数据集,比较IAVOA算法与现有启发式算法对MLP训练后,MLP对数据分类的性能,仿真结果表明,IAVOA算法训练的MLP在数据分类准确率、全局搜索能力、收敛速度和稳定性方面均具有良好的性能。Aiming at the problem that the algorithm is sensitive to initial value,easy to fall into local optimum and slow in convergence when training MLP,an improved algorithm IAVOA was proposed for the heuristic algorithm African vulture optimization algorithm.The logistic chaotic mapping was introduced when initializing the population to increase the diversity of the population.Adaptive coefficients were added to the optimal vultures and sub optimal vultures,and the guiding effects of these two types of vultures on ordinary vultures were automatically adjusted.The improved algorithm IAVOA was used to train MLP.To improve the accuracy of MLP,the average value of mean square error was used as the fitness function to find the best combination of MLP connection weight and deviation.Four classification datasets were selected to compare the performance of MLP for data classification between IAVOA algorithm and existing classical algorithm after MLP training.Simulation results show that the MLP trained using IAVOA algorithm has better performance indicators in data classification,and the improved algorithm has the advantages of strong global search capability,high convergence speed and high convergence accuracy.

关 键 词：优化 分类 非洲秃鹫算法 多层感知器 前馈神经网络 自适应系数 收敛 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]