Agreement-Discrepancy-Selection: Active Learning with Progressive Distribution Alignment

Authors

  • Mengying Fu University of Chinese Academy of Sciences
  • Tianning Yuan University of Chinese Academy of Sciences
  • Fang Wan University of Chinese Academy of Sciences
  • Songcen Xu Noah’s Ark Lab, Huawei Technologies Co., Ltd.
  • Qixiang Ye University of Chinese Academy of Sciences, China

Keywords:

Active Learning

Abstract

In active learning, the ignorance of aligning unlabeled samples' distribution with that of labeled samples hinders the model trained upon labeled samples from selecting informative unlabeled samples. In this paper, we propose an agreement-discrepancy-selection (ADS) approach, and target at unifying distribution alignment with sample selection by introducing adversarial classifiers to the convolutional neural network (CNN). Minimizing classifiers' prediction discrepancy (maximizing prediction agreement) drives learning CNN features to reduce the distribution bias of labeled and unlabeled samples, while maximizing classifiers' discrepancy highlights informative samples. Iterative optimization of agreement and discrepancy loss calibrated with an entropy function drives aligning sample distributions in a progressive fashion for effective active learning. Experiments on image classification and object detection tasks demonstrate that ADS is task-agnostic, while significantly outperforms the previous methods when the labeled sets are small.

Downloads

Published

2021-05-18

How to Cite

Fu, M., Yuan, T., Wan, F., Xu, S., & Ye, Q. (2021). Agreement-Discrepancy-Selection: Active Learning with Progressive Distribution Alignment. Proceedings of the AAAI Conference on Artificial Intelligence, 35(8), 7466-7473. Retrieved from https://ojs.aaai.org/index.php/AAAI/article/view/16915

Issue

Section

AAAI Technical Track on Machine Learning I