Improved Algorithms for Agnostic Pool-based Active Classification

             Julian Katz-Samuels 1 Jifan Zhang 2 Lalit Jain 2 Kevin Jamieson 2

             Abstract               learning methods require that this data be labeled and reli-
    We consider active learning for binary classifi-    able labels are costly since they require human intervention.
    cation in the agnostic pool-based setting. The     The cost of providing labels has become one of the main
    vast majority of works  ...