摘要
Tri-training算法是半监督协同算法里的经典算法,该文针对算法中分类器的使用做了一些改进,由原先单一的分类器换成两个不同分类器的组合。使用SVM分类器和最大熵分类器的不同组合作为Tri-training算法里的三个分类器构成分类器模型,然后分别对稀疏型数据、密集型数据与原始Tri-training算法进行实验比较,从而验证改进的有效性。
Tri-training algorithm is a classical algorithm in semi-supervised learning. In this paper, we have im-proved the use of classifiers by combining two different classifiers instead of only employing one. With the differ-ent combinations of SVM classifier and Maximum Entropy classifier, we formed three classifiers of Tri-training algorithm and shaped the experimental model. Then we compared sparse data and intensive data with the original Tri-training algorithm. The results have confirmed the validity of the improvement.
出处
《苏州科技学院学报(自然科学版)》
CAS
2014年第2期52-56,共5页
Journal of Suzhou University of Science and Technology (Natural Science Edition)
基金
安徽省高校自然科学研究重点项目(KJ2011A048)