Implementation of the $Complex^2$ method proposed in "How to Turn Your Knowledge Graph Embeddings into Generative Models"

This repository implements a heterogenous-graph version of $Complex^2$ method that implicitly constrains edge connections within edge type. This implementation uses PyTorch and is compatible with PyTorch-Geometric data objects.

The method implemented in this repository is based on the work of Loconte et al. (citation below).

@misc{loconte2024turnknowledgegraphembeddings,
title={How to Turn Your Knowledge Graph Embeddings into Generative Models}, author={Lorenzo Loconte and Nicola Di Mauro and Robert Peharz and Antonio Vergari},
year={2024},
eprint={2305.15944},
archivePrefix={arXiv},
primaryClass={cs.LG},
url={https://arxiv.org/abs/2305.15944}, }