quark0 / ANALOGY

Analogical inference for knowledge graph completion

ANALOGY

The repository provides lightweight C++ implementations for the following papers

• Analogical Inference for Multi-Relational Embeddings. Hanxiao Liu, Yuexin Wu and Yiming Yang. ICML 2017.

• Complex Embeddings for Simple Link Prediction. Théo Trouillon, Johannes Welbl, Sebastian Riedel, Éric Gaussier and Guillaume Bouchard. ICML 2016.

• Embedding Entities and Relations for Learning and Inference in Knowledge Bases. Bishan Yang, Wen-tau Yih, Xiaodong He, Jianfeng Gao and Li Deng. ICLR 2015.

Basic Usage

Training

make && ./main -algorithm Analogy -model_path output.model

Prediction

./main -algorithm Analogy -model_path output.model -prediction 1

The program runs with 32 threads by default. For more options, please refer to main.cc.

Contributors

Please cite the following if you use the code for publication

@article{liu2017analogical,
  title={Analogical Inference for Multi-Relational Embeddings},
  author={Liu, Hanxiao and Wu, Yuexin and Yang, Yiming},
  journal={arXiv preprint arXiv:1705.02426},
  year={2017}
}