摘要
针对传统的Adaboost训练算法在训练过程中可能出现训练退化和训练目标类权重分布过适应的问题,提出一种改进的Adaboost训练算法.改进算法通过调整加权误差分布限制目标类权重的扩张,并且最终分类器输出形式以概率值输出代替传统的离散值输出,提高了训练结果的检测率.实验结果表明,改进的Adaboost算法在Inria数据集上取得了较好效果.
In view of the problem of degradation issues as well as the distribution of target class weights adapted to the phenomenon that may arise in the training process of the traditional Adaboost algorithm,the authors introduced a few improved methods to these problems.The article presented a modified Adaboost algorithm based on the adjusted weighted error distribution to limit the expansion weights.In addition,the Adaboost algorithm improved the classifier output forms,i.e.,using output of the probability value instead of the discrete value and increased the detection rate more dramatically.Experiment shows that the test rate of the improved Adaboost algorithm could achieve excellent results in the Inria data set.There are good prospects of application in the field of video security surveillance.
出处
《吉林大学学报(理学版)》
CAS
CSCD
北大核心
2011年第3期498-504,共7页
Journal of Jilin University:Science Edition
基金
国家自然科学基金(批准号:60873147)
国家高技术研究发展计划863项目基金(批准号:2008AA10Z224)
吉林省科技发展计划项目(批准号:20060527)
关键词
误差分布
ADABOOST算法
权重更新
正负误差比
分类器输出
error distribution
Adaboost algorithm
weight update
positive and negative error ratio
classifier output