Awesome

MDE : Multiple Distance Embeddings for Link Prediction in Knowledge Graphs

This is the PyTorch implementation of MDE. The implementation is tailored for cpu servers and performs distributed testing using 8 CPU cores. A GPU version that includes Self-Adversarial Negative Sampling is implemented in here.

** Training ** : To train the model from the command line :

python MDE_Model.py -t task -d dataset_name

Where the task is “train” here and “dataset_name” can be one of WN18, WN18Rr, FB15K, and FB15K237

Or for MDE_NN:

python MDE_NN_Model.py -t task -d dataset_name

For example: python MDE_Model.py -t train -d WN18RR

During the training, a test will be executed after every 50 iterations.

Citation If you use the codes, please cite the following paper:

@inproceedings{sadeghi2020mde,
  title={MDE: Multiple Distance Embeddings for Link Prediction in Knowledge Graphs},
  author={Sadeghi, Afshin and Graux, Damien and Shariat Yazdi, Hamed and Lehmann, Jens},
  booktitle={24th European Conference on Artificial Intelligence, ECAI},
  year={2020},
  url={http://ecai2020.eu/papers/1271_paper.pdf}
}