README.md

THAN

Requirements

Dependencies (with python >= 3.9):

numpy==1.20.3
pandas==1.3.3
torch==1.9.1
scikit_learn==0.24.2

run

pip install -r requirements.txt

Preprocessing

Dataset

Create a folder data to store source data files.

Download dataset from:

• https://grouplens.org/datasets/movielens/100k
• http://snap.stanford.edu/data/higgs-twitter.html
• http://snap.stanford.edu/data/sx-mathoverflow.html

Preprocess the data

If there is no data in processed folder, run

python process_data.py

Training

temporal link prediction task

# on movielens dataset
python learn_edge_all.py -d movielens --bs 500 --n_degree 8 --n_epoch 30 --lr 1e-3 --gpu 0

# on twitter dataset
python learn_edge_all.py -d twitter --bs 800 --n_degree 10 --n_epoch 20 --lr 1e-4 --gpu 0

inductive experiment

python learn_edge_new.py -d movielens --bs 500 --n_degree 8 --n_epoch 30 --lr 1e-3 --gpu 0

# on twitter dataset
python learn_edge_new.py -d twitter --bs 800 --n_degree 10 --n_epoch 20 --lr 1e-4 --gpu 0

If the memory size of your GPU is less than 24GB, please reduce the batch size.