摘要
针对小规模训练样本不足以支持学习器对含有大量潜在不确定因素的未标样本集分类的问题,提出了一种基于信息熵的主动学习方法,引入信息熵的离散事件概率估计理论,通过对未标文档熵值的计算,结合二阶段学习策略,主动学习利用现有知识,结合实验样本环境,主动地选取最有可能的解决问题的样本并标注它们的类别,获得新的参数,重新训练分类器,选择最有利分类器性能的样本,迭代直到未标样本集为空。实验结果表明,该方法取得了较好的分类效果。
Most of supervised machine learning methods led to poor performance when work on limited tagged data. Investigated a novel semi- supervised learning method based on active learning with information entropy. An optimization strategy of selecting part of instances from unlabeled examples for classifying in each iteration, based on active learning from unhbeled examples, was presented. The experiment results show that our method achieve high performance on small tagged data.
出处
《计算机技术与发展》
2010年第2期110-113,共4页
Computer Technology and Development
基金
国家自然科学基金(60673060)
关键词
信息熵
半监督学习
主动学习
分类
information entropy
semi-supervised learning
active learning
elassification