Adaptive-Halting Policy Network for Early Classification

Published in KDD, 2019

This paper proposes the first deep learning solution to Early Classification, which is the multi-objective optimization problem of picking a timestep at which a classification can be made with tunable confidence.

  title={Adaptive-Halting Policy Network for Early Classification},
  author={Hartvigsen, Thomas and Sen, Cansu and Kong, Xiangnan and Rundensteiner, Elke},
  booktitle={Proceedings of the 25th ACM SIGKDD international conference on Knowledge discovery and data mining},