Exploiting Universum data in AdaBoost using gradient descent
作者：
Highlights：
• We address a novel boosting algorithm by taking advantage of Universum data.
• A greedy, stagewise, functional gradient procedure is taken to derive the method.
• Explicit weighting schemes for labeled and Universum samples are provided.
• Practical conditions to verify effectiveness of Universum learning are described.
• This algorithm obtains superior performances over AdaBoost with Universum data.
摘要
•We address a novel boosting algorithm by taking advantage of Universum data.•A greedy, stagewise, functional gradient procedure is taken to derive the method.•Explicit weighting schemes for labeled and Universum samples are provided.•Practical conditions to verify effectiveness of Universum learning are described.•This algorithm obtains superior performances over AdaBoost with Universum data.
论文关键词：AdaBoost,Gradient boost,UAdaBoost,Universum,U-SVM
论文评审过程：Received 1 August 2012, Revised 15 February 2014, Accepted 25 April 2014, Available online 28 May 2014.
论文官网地址：https://doi.org/10.1016/j.imavis.2014.04.009
