Official Python3 implementations of ICML papers On Transportation of Mini-batches: A Hierarchical Approach and Improving Mini-batch Optimal Transport via Partial Transportation.
Details of the model architecture and experimental results can be found in our following papers:
@InProceedings{nguyen2021transportation,
title={On Transportation of Mini-batches: A Hierarchical Approach},
author={Khai Nguyen and Dang Nguyen and Quoc Nguyen and Tung Pham and Hung Bui and Dinh Phung and Trung Le and Nhat Ho},
booktitle={Proceedings of the 39th International Conference on Machine Learning},
year={2022},
}
@InProceedings{nguyen2021improving,
title={Improving Mini-batch Optimal Transport via Partial Transportation},
author={Khai Nguyen and Dang Nguyen and Tung Pham and Nhat Ho},
booktitle = {Proceedings of the 39th International Conference on Machine Learning},
year={2022},
}
Please CITE our papers whenever this repository is used to help produce published results or incorporated into other software.
This implementation is made by Khai Nguyen and Dang Nguyen. README is on updating process.
The code is implemented with Python (3.9) and Pytorch (1.10.1).
To install the required python packages, run
pip install -r requirements.txt
The scalable implementation of the batch of mini-batches scheme and the conventional averaging scheme of mini-batch transportation types: optimal transport (OT), partial optimal transport (POT), unbalanced optimal transport (UOT), and sliced optimal transport for:
- Deep Generative Models
- Deep Domain Adaptation
- Approximate Bayesian Computation
- Color Transfer
- Gradient Flow
The structure of DeepDA is largely based on JUMBOT and ALDA. The structure of ABC is largely based on SlicedABC. We are very grateful for their open sources.