A hybrid algorithm for Bayesian network structure learning with application to multi-label learning
作者：
Highlights：
• A novel hybrid algorithm to learn a local Bayesian network structure around a target variable is proposed.
• Well known benchmarks with various data sizes are first used to assess the quality of the learned structure.
• The ability of this algorithm to solve the multi-label learning problem is then investigated.
• The idea is to learn the independence map of the joint distribution of the label set conditioned on the input features.
• Experiments support the conclusions that local structural learning is well suited to multi-label learning.
摘要
•A novel hybrid algorithm to learn a local Bayesian network structure around a target variable is proposed.•Well known benchmarks with various data sizes are first used to assess the quality of the learned structure.•The ability of this algorithm to solve the multi-label learning problem is then investigated.•The idea is to learn the independence map of the joint distribution of the label set conditioned on the input features.•Experiments support the conclusions that local structural learning is well suited to multi-label learning.
论文关键词：Bayesian networks,Multi-label learning,Markov boundary,Feature subset selection
论文评审过程：Available online 9 May 2014.
论文官网地址：https://doi.org/10.1016/j.eswa.2014.04.032
