This library contains a Tensorflow implementation of the Laten Fact Model and Latent Information model for Gaussian and Von-Mises Fisher latent priors, using the re-parametrisation trick to learn the distributional parameters. The VMF re-parametrisation trick is as presented in [1](http://arxiv.org/abs/1804.00891). Check out the authors of VMF blogpost (https://nicola-decao.github.io/s-vae). The Gaussian re-parametrisation trick is a Tensorflow probability function.

• python>=3.6
• *tf-nightly: https://tensorflow.org
• *tfp-nightly: https://www.tensorflow.org/probability/
• scipy: https://scipy.org

To install, run

$ python setup.py install

• Alexander Cowen-Rivers (GitHub)

• Pasquale Minervini (GitHub)
• Sebastian Riedel (GitHub)

For:

• Models see Latent Fact Model and Latent Information Model
• Paper see ACR

Train variational knowledge graph model, on nations dataset with normal prior using DistMult scoring function :

python main_LIM.py  --no_batches 10 --epsilon 1e-07 --embedding_size 50 --dataset nations --alt_prior False --lr 0.001 --score_func DistMult --negsamples 5 --projection False --distribution normal --file_name /User --s_o False

1. Clone or download this repository.
2. Prepare your data, or use any of the six included KG datasets.

Please cite [1] and [2] in your work when using this library in your experiments.

For questions and comments, feel free to contact ACR(mailto:[email protected]).

MIT

[1] Davidson, T. R., Falorsi, L., De Cao, N., Kipf, T.,
and Tomczak, J. M. (2018). Hyperspherical Variational
Auto-Encoders. arXiv preprint arXiv:1804.00891.

BibTeX format:

@article{s-vae18,
  title={Hyperspherical Variational Auto-Encoders},
  author={Davidson, Tim R. and
          Falorsi, Luca and
          De Cao, Nicola and
          Kipf, Thomas and
          Tomczak, Jakub M.},
  journal={arXiv preprint arXiv:1804.00891},
  year={2018}
}