3lLobo / embed

Implementation of basic KG embedding methods. Meant both as an example implementation and as a testbed for simple improvements.

Variational embedings

Forked from https://github.com/pbloem/embed

Implementation of basic KG embedding method DistMult with the option to do variational sampling on the latent space and learn with bayesian inference.

Running

To run the DistMult LP with the optimal parameters presented in [1] on the FB15k-237 dataset.

python experiments/lp_distmult_fb.py

To run the same experiment with variational sampling.

python experiments/lp_distmult_fb.py --vae

References

[1] @article{winstead2013you, title={You can teach an old dog new tricks: A qualitative analysis of how residents of senior living communities may use the web to overcome spatial and social barriers}, author={Winstead, Vicki and Anderson, William A and Yost, Elizabeth A and Cotten, Shelia R and Warr, Amanda and Berkowsky, Ronald W}, journal={Journal of Applied Gerontology}, volume={32}, number={5}, pages={540--560}, year={2013}, publisher={Sage Publications Sage CA: Los Angeles, CA} }