DySAT: Deep Neural Representation Learning on Dynamic Graphs via Self-Attention Networks

This is a pytorch implementation of DySAT. All codes are adapted from official implementation in TensorFlow. This implementation is only tested using dataset Enron, and the results is inconsistent with official results (better than that). Code review and contribution is welcome!

Raw Data Process

cd raw_data/Enron
pyhton process.py

The processed data will stored at 'data/Enron"

Training

python train --dataset Enron --time_steps 16