Awesome

HIP network

Environment

python 3.6.9
torch==1.7.0+cu101
torch==1.7.0+cu101
torch-scatter==2.0.5
torchvision==0.8.1+cu101
dgl-cu100==0.5.2

Dataset

There are five datasets (from previous RE-NET and CyGNet): ICEWS18, ICEWS14, GDELT, WIKI, and YAGO. These datasets are for the extrapolation problem.

• Times of test set should be larger than times of train and valid sets.
• Times of valid set also should be larger than times of train set.

Each data folder has 'stat.txt', 'train.txt', 'valid.txt', 'test.txt'

• 'stat.txt': First value is the number of entities, and second value is the number of relations.
• 'train.txt', 'valid.txt', 'test.txt': First column is subject entities, second column is relations, and third column is object entities. The fourth column is time. The fifth column is ignored in our experiment.

Run the main experiment

Train the model and test. python train.py -d ICEWS18 --gpu 2 --dropout 0.5 --n-hidden 200 --lr 1e-3 --max-epochs 100 --batch-size 1024 --valid-every 10 --test-every 2 For other datasets, the only thing need to do is replacing the ICEWS18 with other names (YAGO, WIKI, ICEWS14, GDELT).

Other parameters

You can find more details at train.py.

Train time

• All models achieve their best results within 100 epochs.
• Since we cached the information for each time step, an epoch takes one to ten minutes on NVIDIA TITAN RTX Graphics Processing Units.

solid seed

In train.py:

• seed = 999
• np.random.seed(seed)
• torch.manual_seed(seed)
• torch.cuda.manual_seed_all(seed)

reference

@inproceedings{DBLP:conf/ijcai/HeZL0ZZ21,  
  author    = {Yongquan He and  
               Peng Zhang and  
               Luchen Liu and  
               Qi Liang and  
               Wenyuan Zhang and  
               Chuang Zhang},  
  title     = {HIP Network: Historical Information Passing Network for Extrapolation  
               Reasoning on Temporal Knowledge Graph},  
  booktitle = {Proceedings of the Thirtieth International Joint Conference on Artificial
               Intelligence, IJCAI 2021, Virtual Event / Montreal, Canada, 19-27
               August 2021},  
  pages     = {1915--1921},  
  year      = {2021},  
  url       = {https://doi.org/10.24963/ijcai.2021/264}  
}