mnick / holographic-embeddings

Code for experiments in the AAAI 2016 paper "Holographic Embeddings of Knowledge Graphs"

Holographic Embeddings of Knowledge Graphs

This repository holds the code for experiments in the paper

Holographic Embeddings of Knowledge Graphs
Maximilian Nickel, Lorenzo Rosasco, Tomaso Poggio, AAAI 2016.

Install

To run the experiments, first install scikit-kge, An open-source python library to compute various knowledge graph embeddings including

• Holographic Embeddings (HolE)
• RESCAL
• TransE
• TransR
• ER-MLP

After scikit-kge is installed, simply clone this repository via

git clone [e-mail]:mnick/holographic-embeddings.git

and run the experiments as detailed in the next section

Experiments

The repository holds scripts of the form

run_<model>_<dataset>.sh

which runs the experiments for dataset with the best parameters for model.

The full code for the experiments can be found in the kg and countries subfolders. The python scripts in these subfolders should be easy to use for grid search.