Installation

Create a conda environment with pytorch and scikit-learn :

conda create --name tcom_env python=3.8
source activate tcom_env
conda install --file requirements.txt -c pytorch

Datasets

python process_icews.py

python process_gdelt.py

This will create the files required to compute the filtered metrics.

Reproducing results of TCompoundE

In order to reproduce the results of TCompoundE on the four datasets in the paper, run the following commands

python  learner.py --dataset ICEWS14 --emb_reg 0.01 --time_reg 0.01 --learning_rate 0.01  --rank 6000  --batch_size 4000  --max_epochs 400

python  learner.py --dataset ICEWS05-15 --emb_reg 0.002   --time_reg 0.1  --learning_rate 0.08  --rank 8000  --batch_size 6000  --max_epochs 100

python  learner.py --dataset GDELT --emb_reg 0.001   --time_reg 0.001  --learning_rate 0.35  --rank 6000  --batch_size 2000  --max_epochs 50