https://github.com/ibalazevic/HypER

Hypernetwork Knowledge Graph Embeddings
https://github.com/ibalazevic/HypER

Last synced: about 1 month ago
JSON representation

Hypernetwork Knowledge Graph Embeddings

• Host: GitHub
• URL: https://github.com/ibalazevic/HypER
• Owner: ibalazevic
• License: mit
• Created: 2018-05-16T19:22:41.000Z (over 6 years ago)
• Default Branch: master
• Last Pushed: 2019-09-21T08:16:58.000Z (about 5 years ago)
• Last Synced: 2024-08-01T22:42:04.078Z (4 months ago)
• Language: Python
• Homepage:
• Size: 13.2 MB
• Stars: 53
• Watchers: 6
• Forks: 9
• Open Issues: 5
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

• awesome-network-embedding - [Python PyTorch

README

        

## HypER: Hypernetwork Knowledge Graph Embeddings



  



This codebase contains PyTorch implementation of the paper:

> Hypernetwork Knowledge Graph Embeddings.

> Ivana Balažević, Carl Allen, and Timothy M. Hospedales.

> International Conference on Artificial Neural Networks, 2019.

> [[Paper]](https://arxiv.org/pdf/1808.07018.pdf)

### Running a model

To run the model, execute the following command:

`CUDA_VISIBLE_DEVICES=0 python hyper.py --algorithm HypER --dataset FB15k-237`

Available algorithms are:

    HypER

    HypE

    DistMult

    ComplEx

    ConvE

Available datasets are:

    

    FB15k-237

    WN18RR

    FB15k

    WN18

To reproduce the results from the paper, use the following combinations of hyperparameters with `batch_size=128`, `ent_vec_dim=200` and `rel_vec_dim=200`:

dataset | lr | dr | input_dropout | feature_map_dropout | hidden_dropout | label_smoothing 

:--- | :---: | :---: | :---: | :---: | :---: | :---: | 

FB15k | 0.005 | 0.995 | 0.2 | 0.2 | 0.3 | 0.

WN18 | 0.001 | 1.0 | 0.2 | 0.2 | 0.3 | 0.1

FB15k-237 | 0.0001 | 0.995 | 0.3 | 0.2 | 0.3 | 0.1

WN18RR | 0.005 | 1.0 | 0.2 | 0.2 | 0.3| 0.1

### Requirements

The codebase is implemented in Python 3.6.6. Required packages are:

    numpy      1.14.5

    pytorch    0.4.0

### Citation

If you found this codebase useful, please cite:

    @inproceedings{balazevic2019hypernetwork,

    title={Hypernetwork Knowledge Graph Embeddings},

    author={Bala\v{z}evi\'c, Ivana and Allen, Carl and Hospedales, Timothy M},

    booktitle={International Conference on Artificial Neural Networks},

    year={2019}

    }