Learning Bayesian network classifiers from label proportions
作者：
Highlights：
• A Structural-EM method to learn Bayesian network classifiers for the LLP problem.
• Variants of the method designed to deal with very-complex LLP scenarios.
• Only (joint) label assignments that fulfill the LP of the groups are considered.
• A framework for testing LLP methods that covers the spectrum of LLP complexities.
• Good performance behavior in different experimental settings.
摘要
Highlights•A Structural-EM method to learn Bayesian network classifiers for the LLP problem.•Variants of the method designed to deal with very-complex LLP scenarios.•Only (joint) label assignments that fulfill the LP of the groups are considered.•A framework for testing LLP methods that covers the spectrum of LLP complexities.•Good performance behavior in different experimental settings.
论文关键词：Supervised classification,Learning from label proportions,Structural EM algorithm,Bayesian network classifiers
论文评审过程：Received 27 June 2012, Revised 25 April 2013, Accepted 2 May 2013, Available online 15 May 2013.
论文官网地址：https://doi.org/10.1016/j.patcog.2013.05.002
