Bridging OOD Generalization and Detection: A Graph-Theoretic View

This is the source code accompanying the NeurIPS 2024 paper of Bridging OOD Generalization and Detection: A Graph-Theoretic View by Han Wang, Yixuan Li.

Environment Preparation

You can set up the environment by running the following commands:

conda create --name graph_ood python==3.9.17
conda activate graph_ood
pip install -r requirements.txt

Dataset Preparation

Download the dataset to the folder ./data

Here are the links for the semantic OOD datasets used in this paper: SVHN, Textures, Places365, LSUN, LSUN-R, and iSUN. We refer readers to these links Scone and Robustness to download the covariate OOD datasets.

For large-scale experiments, we use iNaturalist as the semantic OOD dataset. The dataset can be downloaded via the following link:

wget http://pages.cs.wisc.edu/~huangrui/imagenet_ood_dataset/iNaturalist.tar.gz

Demo

To pretrain the model (e.g., using the CIFAR-10-C as the covariate OOD dataset and SVHN as the semantic OOD dataset), you can modify the variable checkpoints_dir in the ./CIFAR/run_pt.sh, and run the following command:

sh run_pt.sh graph_ood cifar10 svhn svhn

pi_1 is set to 0.5 and pi_2 is set to 0.1 as default. See the run_pt.sh for more details and options.

To finetune the model, you can modify the variable load_pretrained in the ./CIFAR/run_ft.sh and run the following command:

sh run_ft.sh graph_ood cifar10 svhn svhn

Checkpoints

We also provide some checkpoints during the pretraining process, please see ./CIFAR/snapshots.

Citations

If you find our work useful, please consider citing our paper:

@article{wang2024bridging,
  title={Bridging OOD Detection and Generalization: A Graph-Theoretic View},
  author={Wang, Han and Li, Yixuan},
  journal={Advances in Neural Information Processing Systems},
  year={2024}
}

Our codebase is heavily based on the followings:

@inproceedings{
    sun2023sorl,
    title={A Graph-Theoretic Framework for Understanding Open-World Semi-Supervised Learning},
    author={Yiyou Sun and Zhenmei Shi and Yixuan Li},
    booktitle={Advances in Neural Information Processing Systems},
    year={2023},
    url={https://openreview.net/forum?id=ZITOHWeAy7}
}

@inproceedings{bai2023feed,
      title={Feed Two Birds with One Scone: Exploiting Wild Data for Both Out-of-Distribution Generalization and Detection}, 
      author={Haoyue Bai and Gregory Canal and Xuefeng Du and Jeongyeol Kwon and Robert D Nowak and Yixuan Li},
      booktitle = {International Conference on Machine Learning},
      year = {2023}
}

@inproceedings{katz2022training,
  title={Training Ood Detectors in Their Natural Habitats},
  author={Katz-Samuels, Julian and Nakhleh, Julia B and Nowak, Robert and Li, Yixuan},
  booktitle={International Conference on Machine Learning},
  year={2022}
}