CompoundE

Code for "CompoundE: Knowledge Graph Embedding with Translation, Rotation and Scaling Compound Operations"

OGB installation

https://github.com/snap-stanford/ogb

Recommended order of installation

• Create a python 3.8.8 environment using Anaconda
• Install torch 1.11.0 using Anaconda
• Install torch-geometric using Anaconda
• Install ogb using pip
• Install other packages using pip

After installing the OGB environment, to reproduce the results

CUDA_VISIBLE_DEVICES=0 python run.py --do_train --cuda --do_valid --do_test --evaluate_train \
            --model CompoundE -n 250 -b 4096 -d 100 -g 7 -a 1.0 -adv -tr \
            -lr 0.005 --max_steps 300000 --cpu_num 8 --test_batch_size 32  --print_on_screen

Acknowledgement

We refer to the code of PairRE. Thanks for their contributions.

Citation

If you find the source codes useful, please consider citing our paper:

@inproceedings{ge2023compounding,
  title={Compounding geometric operations for knowledge graph completion},
  author={Ge, Xiou and Wang, Yun Cheng and Wang, Bin and Kuo, C-C Jay},
  booktitle={Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  pages={6947--6965},
  year={2023}
}