In this paper, we propose a novel dynamic graph representation learning framework featuring Temporal Anonymous Walks with Restart, Memory Augmentation and Neighbor Co-occurrence.
Install packages in requirements.txt (with python >= 3.9):
The datasets come from Towards Better Evaluation for Dynamic Link Prediction,
which can be downloaded here.
Please download them, and store their csv files in a folder named
data/.
We use the npy format to save the features in binary format. If edge features or nodes
features are absent, they will be replaced by a vector of zeros.
For example:
python utils/preprocess_data.py --data wikipedia --bipartite
Dynamic link prediction task:
python train_link_prediction.py --negative_sampling_strategy random --use_memory --enable_walk --enable_dynamic_restart --enable_neighbor_cooc --n_runs 5
Dynamic node classification (this requires a trained model from the link prediction task):
python train_node_classification.py --use_memory --enable_walk --enable_dynamic_restart --enable_neighbor_cooc --n_runs 5 --model_date 20241206102940
This repository accompanies our paper accepted at The Web Conference (TheWebConf) 2026.
Please cite the following paper:
@inproceedings{farokhi2026tawrmac,
author = {Farokhi, Soheila and Qi, Xiaojun and Karimi, Hamid},
title = {TAWRMAC: A Novel Dynamic Graph Representation Learning Method},
year = {2026},
isbn = {9798400723070},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
url = {https://doi.org/10.1145/3774904.3792163},
doi = {10.1145/3774904.3792163},
booktitle = {Proceedings of the ACM Web Conference 2026},
pages = {684–695},
numpages = {12},
keywords = {dynamic graph representation learning, graph neural network, temporal walks, link prediction, node classification},
location = {United Arab Emirates},
series = {WWW '26}
}
This project is licensed under the MIT License – see the LICENSE file for details.