README.md

Code for PseudoCal@ICML 2024

Pseudo-Calibration: Improving Predictive Uncertainty Estimation in Unsupervised Domain Adaptation

Prerequisites

• python == 3.7.13
• cudatoolkit == 10.1.243
• pytorch ==1.7.1
• torchvision == 0.8.2
• numpy, scikit-learn, PIL, argparse

Demo

• Configure the PyTorch environment.
• Download the Office-Home dataset. Configure the data lists in data and the checkpoints in logs.
• Run the code in pseudocal.sh.

Citation

@inproceedings{hu2024pseudocalibration,
    title={Pseudo-Calibration: Improving Predictive Uncertainty Estimation in Unsupervised Domain Adaptation},
    author={Dapeng Hu and Jian Liang and Xinchao Wang and Chuan-Sheng Foo},
    booktitle={Forty-first International Conference on Machine Learning},
    year={2024}
}

Contact

• [email protected]

Credit

• The code is heavily borrowed from TransCal.