This code repository implements Open-world Domain Adapted Classification (ODAC). Under the general setting, where we have both an open-world nature, and a distribution shift, we enforce our model to be cautiously confident, i.e. to output more confident predictions for instances of known classes, but not for those of unknown classes.

If you find this work useful, please cite:

@inproceedings{srey2024open,
  title={Open-world learning under dataset shift},
  author={Srey, Ponhvoan and Zhang, Yuhui and Kanamori, Takafumi},
  booktitle={2024 IEEE Conference on Artificial Intelligence (CAI)},
  pages={1040--1042},
  year={2024},
  organization={IEEE}
}