On the Power of Localized Perceptron for Label-Optimal Learning of
            Halfspaces with Adversarial Noise

                               Jie Shen 1

             Abstract                  the learner must pay for each label it wishes to be revealed.
   We study online active learning of homogeneous         The goal is to design querying strategies to avoid less infor-
   halfspaces in Rd with adversarial noise where         mative labeling requests, e.g. the labels that can be inferred
  ...