摘要
针对基于PSO的AdaBoost算法(PSO-AdaBoost)的不足,分析了传统目标函数不能适应多个弱分类器拥有相同最小错误率时弱分类器的选择问题,提出了解决这一问题的有效方法。新方法使用特征值和阈值的绝对值差衡量错分样本的错误程度,结合相对熵理论形成PSO算法的适应度函数,使其根据错分样本的错误程度挑选最佳弱分类器。实验结果表明,所提算法具有较高的检测率和较小的泛化错误。
Focusing on the disadvantage of the AdaBoost algorithm based on PSO,this paper mainly analyzed the issue that the traditional target function could not adapt to the problem of weak classifiers selection when they had the same minimum error rate and a new method was advanced to avoid the problem.The new method used the absolute difference between the threshold and feature to measure the extent of misclassification and combined with the relative entropy principle as the fitness function.In this way,the new fitness function could select the best weak classifiers more accuracy.Experimental results indicate that the method can achieve both better performance and less generalization error.
出处
《计算机应用研究》
CSCD
北大核心
2012年第1期127-129,共3页
Application Research of Computers
基金
甘肃省教育厅研究生导师基金资助项目(1014ZTC089)
甘肃省财政厅科研项目(0914ZTB148)
关键词
人脸检测
粒子群优化
ADABOOST算法
相对熵
训练算法
face detection
particle swarm optimization（PSO）
AdaBoost algorithm
relative entropy
training algorithm
