Multi-class boosting with asymmetric binary weak-learners
作者：
Highlights：
• Multi-class boosting procedure with binary weak-learners.
• Novel multi-class vectorial encoding producing different margin values.
• Margins depend on the asymmetry of the problem posed to the weak-learner.
• Provides statistically significant improvements in performance.
• Opens research venues in multi-{class,label,dimensional} classification.
摘要
Highlights•Multi-class boosting procedure with binary weak-learners.•Novel multi-class vectorial encoding producing different margin values.•Margins depend on the asymmetry of the problem posed to the weak-learner.•Provides statistically significant improvements in performance.•Opens research venues in multi-{class,label,dimensional} classification.
论文关键词：AdaBoost,Multi-class classification,Asymmetric binary weak-learners,Class imbalance
论文评审过程：Received 9 May 2012, Revised 11 November 2013, Accepted 23 November 2013, Available online 2 December 2013.
论文官网地址：https://doi.org/10.1016/j.patcog.2013.11.024
