Self-Training Associative Classification
DEMAND-DRIVEN ASSOCIATIVE CLASSIFICATION(2011)
摘要
The acquisition of training examples usually requires skilled human annotators to manually label the relationship between inputs and outputs. Due to various reasons, annotators may face inputs that are hard to label (Chapelle et al. Semi-supervised learning. MIT Press, Cambridge, 2006). The cost associated with this labeling process thus may render vast amounts of training examples unfeasible. The acquisition of unlabeled inputs (i.e., inputs for which the corresponding output is unknown), on the other hand, is relatively inexpensive. However, it is worthwhile to label at least some inputs, provided that this effort will be then rewarded with an improvement in classification performance. In this chapter demand-driven associative classification will be extended, so that the corresponding algorithm achieves high classification performance even in the case of limited labeling efforts.
更多查看译文
关键词
Labeling effort,Unlabeled data,Semi-supervised learning,Reliability,Lack of evidence,Name disambiguation
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络