Hybrid learning of Bayesian multinets for binary classification

作者：

Highlights：

• New scoring criterion for hybrid learning of two-component Bayesian multinets.

• The score is suitable when data is split in foreground and background.

• Outperformed state-of-the-art classifiers in different domains of application.

摘要

Highlights•New scoring criterion for hybrid learning of two-component Bayesian multinets.•The score is suitable when data is split in foreground and background.•Outperformed state-of-the-art classifiers in different domains of application.

论文关键词：Conditional log-likelihood,Approximation,Hybrid learning,Discriminative learning,Bayesian networks,Mixtures,Multinets

论文评审过程：Received 18 June 2013, Revised 12 March 2014, Accepted 18 March 2014, Available online 2 April 2014.

论文官网地址：https://doi.org/10.1016/j.patcog.2014.03.019