Paper: Amalgamating Knowledge from Heterogeneous Graph Neural Networks, CVPR'21
See requirements file for more information about how to install the dependencies.
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.
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".
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!
AmalgamateGNN is released under the MIT license. Please see the LICENSE file for more information.