Skip to content

ycjing/AmalgamateGNN.PyTorch

Repository files navigation

PyTorch implementation of AmalgamateGNN

Paper: Amalgamating Knowledge from Heterogeneous Graph Neural Networks, CVPR'21

Method Overview

Dependencies

See requirements file for more information about how to install the dependencies.

Usage

1. Prerequisites

A pool of pre-trained teacher models for knowledge amalgamation. We provide in the folder "teacher_models" two example teacher models, which are pre-trained by using two subsets of PPI, termed as PPI Set1 and PPI Set2, with 60 and 61 biological labels, respectively.

2. Training and Evaluation

Use ka_train_ppi_student.py to train a multi-talented student GNN model. Run python ka_train_ppi_student.py -h to view all the possible parameters.

Example usage:

$ python ka_train_ppi_student.py

Our log file is provided in the folder "logs".

Citation

If you find this code useful for your research, please consider citing:

@inproceedings{jing2021amalgamate,
  title={Amalgamating Knowledge from Heterogeneous Graph Neural Networks},
  author={Jing, Yongcheng and Yang, Yiding and Wang, Xinchao and Song, Mingli and Tao, Dacheng},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
  year={2021}
}

Thanks!

License

AmalgamateGNN is released under the MIT license. Please see the LICENSE file for more information.

About

PyTorch implementation of AmalgamateGNN (CVPR'21)

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages