Active Learning from Imperfect Labelers

@inproceedings{Yan2016ActiveLF,
  title={Active Learning from Imperfect Labelers},
  author={Songbai Yan and Kamalika Chaudhuri and Tara Javidi},
  booktitle={NIPS},
  year={2016}
}
We study active learning where the labeler can not only return incorrect labels but also abstain from labeling. We consider different noise and abstention conditions of the labeler. We propose an algorithm which utilizes abstention responses, and analyze its statistical consistency and query complexity under fairly natural assumptions on the noise and abstention rate of the labeler. This algorithm is adaptive in a sense that it can automatically request less queries with a more informed or less… CONTINUE READING
3
Twitter Mentions

Figures and Topics from this paper.

Citations

Publications citing this paper.
SHOWING 1-10 OF 12 CITATIONS

Active Learning and CSI Acquisition for mmWave Initial Alignment

  • IEEE Journal on Selected Areas in Communications
  • 2019
VIEW 1 EXCERPT
CITES BACKGROUND

Clash Royale Challenge: How to Select Training Decks for Win-rate Prediction

  • 2019 Federated Conference on Computer Science and Information Systems (FedCSIS)
  • 2019
VIEW 2 EXCERPTS
CITES BACKGROUND

Reliable Label-Efficient Learning for Biomedical Image Recognition

  • IEEE Transactions on Biomedical Engineering
  • 2018
VIEW 1 EXCERPT
CITES BACKGROUND

References

Publications referenced by this paper.
SHOWING 1-10 OF 25 REFERENCES

Active learning from noisy and abstention feedback

  • 2015 53rd Annual Allerton Conference on Communication, Control, and Computing (Allerton)
  • 2015
VIEW 10 EXCERPTS

Minimax bounds for acti ve learning.IEEE Transactions on Information Theory

Rui M. Castro, Robert D. Nowak
  • 2008
VIEW 9 EXCERPTS
HIGHLY INFLUENTIAL

Sequential no nparametric testing with the law of the iterated logarithm

Aaditya Ramdas, Akshay Balsubramani
  • InProceedings of the Conference on Uncertainty in Artificial I ntelligence,
  • 2016
VIEW 2 EXCERPTS