HyperKG: Hyperbolic Knowledge Graph Embeddings for Knowledge Base Completion

This repository contains our implementation of the HyperKG: Hyperbolic Knowledge Graph Embeddings for Knowledge Base Completion.

License

This code is partially based on code from the following repositories:

• OpenKE-PyTorch (old)
• poincare-embeddings

Every source code file written exclusively by the author of this repo is licensed under Apache License Version 2.0. For more information, please refer to the license.

Instructions for running:

• Prerequisites :

  • Python, C, C++.
  • Python Libraries: NumPy, SciPy, pytorch (with CUDA support).

• Run the code:

  1. Compile the C, C++ code using: sh make.sh
  2. To analyze HyperKG's performance on a dataset, please run:

     python example_train_poincare.py
     

     All parameters/hyperparameters can be altered by directly modifying the example_train_poincare.py file.
  3. Known issues: There is a portability issue with the original C code provided by OpenKE-PyTorch (old). As a quick workaround, I added a Datatype control variable in the original Config Class. If a segmentation fault occurs after Step 2, then this command con.set_int_type('int64') should be commented out in both example_train_poincare.py and example_test_poincare.py files.

Saved Models:

The folder res/saved_models contains saved models for the experiments WN18RR and FB15k-237.

Contact:

• prodromos DOT kolyvakis AT epfl DOT ch