On Investigating the Conservative Property of Score-Based Generative Models

This repository contains the code implementation of the experiments presented in the paper On Investigating the Conservative Property of Score-Based Generative Models.

The project page is available at: https://chen-hao-chao.github.io/qcsbm/

Directory Structure

• Use the code in qcsbm/gaussian_example to reproduce the experimental results presented in Section 3.1.
• Use the code in qcsbm/2d_examples to reproduce the experimental results presented in Section 3.2.
• Use the code in qcsbm/real_world to reproduce the experimental results presented in Section 5.
• Use the code in qcsbm/autoencoder_example to reproduce the experimental results presented in Section 6.

Dependencies

(Optional) Launch a docker container:

# assume the current directory is the root of this repository
docker run --rm -it --gpus all --ipc=host -v$(pwd):/app nvcr.io/nvidia/pytorch:20.12-py3
# inside the docker container, run:
cd /app

Install the necessary Python packages through the following commands:

pip install -r requirements.txt --use-feature=2020-resolver

Citing QCSBM

If you find this code useful, please consider citing our paper.

@inproceedings{chao2023investigating,
      title={On Investigating the Conservative Property of Score-Based Generative Models}, 
      author={Chen-Hao Chao and Wei-Fang Sun and Bo-Wun Cheng and Chun-Yi Lee},
      year={2023},
      booktitle={International Conference on Machine Learning (ICML)},
}

License

To maintain reproducibility, we freezed the following repository and list its license below:

• yang-song/score_sde_pytorch (at commit 1618dde) is licensed under the Apache-2.0 License

Further changes based on the repository above are licensed under the Apache-2.0 License.