TransH

Introduction

[paper]

Title: Knowledge Graph Embedding by Translating on Hyperplanes

Authors: Zhen Wang, Jianwen Zhang, Jianlin Feng, Zheng Chen

Abstract: TransH improves upon TransE by modeling relations as translations on relation-specific hyperplanes. Each relation has a normal vector defining a hyperplane, and entities are projected onto this hyperplane before the translation. This allows TransH to better model 1-to-N, N-to-1, and N-to-N relations that TransE struggles with.

Running with hopwise

Model Hyper-Parameters:

  • embedding_size (int) : The embedding size of entities and relations. Defaults to 64.

  • margin (float) : The margin used in the TripletMarginLoss. Defaults to 1.0.

A Running Example:

Write the following code to a python file, such as run.py

from hopwise.quick_start import run_hopwise

run_hopwise(model='TransH', dataset='ml-100k')

And then:

python run.py

Tuning Hyper Parameters

If you want to use HyperTuning to tune hyper parameters of this model, you can copy the following settings and name it as hyper.test.

learning_rate choice [0.01,0.005,0.001,0.0005,0.0001]
embedding_size choice [32,64,128]
margin choice [0.5,1.0,2.0]

Note that we just provide these hyper parameter ranges for reference only, and we can not guarantee that they are the optimal range of this model.

Then, with the source code of hopwise (you can download it from GitHub), you can run the run_hyper.py to tuning:

hopwise tune --config_files=[config_files_path] --params_file=hyper.test

For more details about Parameter Tuning, refer to Parameter Tuning.