多标记学习201108-153指南.pdf

中国科技论文在线 多标记学习：问题、算法与数据# 张敏灵* （东南大学计算机科学与工程学院，南京 210096） 5 摘要：在多标记学习框架下，每个样本由单个示例进行表示但同时隶属于多个概念标记。多 标记学习系统的目标是分析概念标记集合已知的训练样本，从而尽可能正确地预测未知样本 的概念标记集合。目前有关多标记学习已经取得了大量的研究成果，本文将对其研究现状进 行简要介绍。首先给出多标记学习的形式化定义及其评价指标，然后介绍已有的多标记学习 算法以及该领域内常用的实验数据集，最后对多标记学习有待进一步研究的问题进行讨论。 10 关键词：人工智能；机器学习；监督学习；多标记学习；泛化 中图分类号：TP181 Learning from Multi-Label Data: Problems, Algorithms and Data Sets 15 ZHANG Minling (School of Computer Science and Engineering, Southeast University, NanJing 210096) Abstract: In multi-label learning, each example is represented by a single instance while associated with multiple class labels. The task of multi-label learning is to correctly predict the label set of unseen example, through analyzing training examples with known label sets. A variety 20 of research results have been achieved on the area of multi-label learning. The state-of-the-art of multi-label learning is briefly reviewed in this paper. Firstly, the formal definition of multi-label learning and its several evaluation metrics are given. After that, existing multi-label learning algorithms and the popular multi-label learning data sets are introduced. Finally, future works on multi-label learning are discussed. 25 Keywords: artificial intelligence; machine learning; supervised learning; multi-label learning; generalization 0 引言 [1] 多标记学习起源于文档分类 研究中遇到的歧义性问题，在该学习框架下，每个对象由 30 单个示例（属性向量）