Publication

An Effective Baseline for Robustness to Distributional Shift

Sushil Thapa December 13, 2021 1 min read 20th IEEE International Conference on Machine Learning and Applications (ICMLA 2021)

This approach adds an abstention class to a classifier. Training combines labeled in-distribution examples with an uncurated set of out-of-distribution examples assigned to that class. The model learns to distinguish familiar categories from inputs on which it should abstain.

Paper (arXiv) · IEEE Xplore

BibTeX

@inproceedings{thulasidasan2021effective,
  title     = {An Effective Baseline for Robustness to Distributional Shift},
  author    = {Thulasidasan, Sunil and Thapa, Sushil and Dhaubhadel, Sayera and Chennupati, Gopinath and Bhattacharya, Tanmoy and Bilmes, Jeff},
  booktitle = {2021 20th IEEE International Conference on Machine Learning and Applications (ICMLA)},
  pages     = {278--285},
  year      = {2021},
  organization = {IEEE},
  doi       = {10.1109/ICMLA52953.2021.00050}
}