Embedding

This Project is contributed by Xiao Han in Tsinghua University.

Datasets

• http://www.ibookman.net
• http://ciia.cs.tsinghua.edu.cn

Supported Papers

• ManifoldE (IJCAI.2016): http://www.ibookman.net/IJCAI.2016.ManifoldE.pdf
• TransG (ACL.2016): http://www.ibookman.net/ACL.2016.TransG.pdf
• SSP (AAAI.2017): http://www.ibookman.net/AAAI.2017.SSP.pdf
• TransA (Arxiv): http://www.ibookman.net/Arxiv.TransA.pdf
• KSR (submitting to ACL.2017): http://www.ibookman.net/Arixv.KSR.pdf

Citation

Conventionally, if this project helps you, please cite our paper, corresponddingly.

• Han Xiao, Minlie Huang, Xiaoyan Zhu. From One Point to A Manifold: Orbit Models for Knowledge Graph Embedding. The 25th International Joint Conference on Artificial Intelligence (IJCAI'16).
• Han Xiao, Minlie Huang, Xiaoyan Zhu. TransG: A Generative Mixture Model for Knowledge Graph Embedding. The 54th Annual Meeting of the Association for Computational Linguistics (ACL'2016).
• Han Xiao, Minlie Huang, Lian Meng, Xiaoyan Zhu. SSP: Semantic Space Projection for Knowledge Graph Embedding with Text Descriptions. The Thirty-First AAAI Conference on Artificial Intelligence (AAAI'17).

Dependency

• Armadillo
  • Armadillo is a high quality linear algebra library (matrix maths) for the C++ language, aiming towards a good balance between speed and ease of use
  • I bet you could master it, just by scanning the examples.
  • Download URL: http://arma.sourceforge.net/download.html
  • What all you should do is to copy the headers into your environment.
• Boost
  • C++ Standard Extensive Library.
  • Download URL:http://www.boost.org/users/download/
  • What all you should do is to copy the headers into your environment. Certainly, you could compile the code just as explained in the website.
• MKL
  • Not Necessary, but I strongly suggest you could take advantage of your devices.

Basic Configuration

• Windows

  • This project is naturally built on Visual Studio 2013 with Intel C++ Compiler 2016. If we share the same development perference, I guess you could start your work, right now.
  • When you decide to compile it with MSC, there is a little trouble, because you shoud adjust your configuration.

• Linux / MAC

  • I also apply the Intel C++ Compiler, which could be substituted by GCC, theoretically.
  • icc -std=c++11 -O3 -xHost -qopenmp -m32 Embedding.cpp

Start

• To justify your data source, please modify the MultiChannelEmbedding\DetailedConfig.hpp.

• To explore the correspondding method, just fill the template in MultiChannelEmbedding\Embedding.cpp with hyper-parameters.

  • model = new MFactorE(FB15K, LinkPredictionTail, report_path, 10, 0.01, 0.1, 0.01, 10);
  • model->run(10000);
  • model->test();
  • delete model;

• Notably, our code needs a little more turns to converge, we suggest 10,000 rounds for each experiment. This is a critical trick for repeating our experiments.

Alias

• OrbitE = ManifoldE
• MFactorE = KSR