Residual Label Propagation Algorithm (ResLPA) approximates residuals between propagated labels with node feature information. Those approximated residuals are added to labels as corrections to improve the accuracy of vanilla LPA.
In experiments on five real-world graphs, ResLPA matches SotA LPA-based methods, GCN-based methods and their integrations.
My presentation of ResLPA on conference ICCAI2021
Install dependencies torch and DGL:
pip3 install -r requirements.txtRun the script with arguments algorithm_name, dataset_name and whether to split the dataset into 60%/20%/20%.
The following command experiments Fast ResLPA on Cora dataset using the default split (a training set containing 20 nodes for each class, a validation set of 500 nodes and a test set of 1000 nodes):
python3 main.py fastreslpa cora falseThis command runs the baseline C&S on Coauthor CS dataset split into 60% as the training set, 20% as the validation set and 20% as the test set.
python3 main.py cs coauthor-cs true- available algorithms: mlp / lpa / cs / gcn / sage / gat / adalpa / gcnlpa / reslpa / fastreslpa
- available datasets: cora / citeseer / pubmed / coauthor-cs / coauthor-phy
@inproceedings{DBLP:conf/iccai/LuoHCZ21,
author = {Yi Luo and
Rui Huang and
Aiguo Chen and
Xi Zeng},
title = {ResLPA: Label Propagation With Residual Learning},
booktitle = {{ICCAI} '21: 2021 7th International Conference on Computing and Artificial
Intelligence, Tianjin China, April 23 - 26, 2021},
pages = {296--301},
publisher = {{ACM}},
year = {2021},
url = {https://doi.org/10.1145/3467707.3467752},
doi = {10.1145/3467707.3467752},
timestamp = {Tue, 28 Sep 2021 15:58:51 +0200},
biburl = {https://dblp.org/rec/conf/iccai/LuoHCZ21.bib},
bibsource = {dblp computer science bibliography, https://dblp.org}
}