snaildm / pyke-kge

A Physical Embedding Model for Knowledge Graphs

A Physical Embedding Model for Knowledge Graphs

This repository contains the implementation of our approach for learning continous vector representation of knowledge graphs.

• Our approach (PYKE) is a physical embedding model for learning embeddings of RDF graphs. By virtue of being akin to a physical simulation, PYKE retains a linear space complexity while generating high quality embeddings. We evaluated our approach with two benchmark datasets and showed that it outperforms state-of-the-art approaches on all tasks, while being close to linear in its time complexity on large KGs.

Installation

python --version
Python 3.6.4
git clone https://github.com/pyke-KGE/pyke-KGE.git
pip install -r requirements.txt

Reproducing reported results

• PYKE Drugbank and PYKE DBpedia notebooks elucidates the workflow of PYKE as well as reproduces the results of type prediction and cluster purity evaluations.