E cient Active Learning

Abstract

We present and analyze an active learning algorithm that is theoretically sound in an agnostic setting, empirically e ective, and as e cient as standard online learning algorithms. This allows us to soundly and effectively optimize the explore/exploit tradeo in active learning at a scale of 10 examples/second. The present work is primarily based on… (More)

2 Figures and Tables

Topics

  • Presentations referencing similar topics