Knowledge Graph Embedding

A TensorFlow-based implementation of knowledge graph embedding models.
Document available here: https://knowledge-graph-embedding.readthedocs.io/en/latest/index.html

Todos

• finish docs
• unit test
• model saving
• early stopping with ranking metric (for now using validation loss)
• reproducible paper experiment

Models

Including following knowledge graph embedding model:

Translating Based

• Unstructured Model (UM)
• Structured Embedding (SE)
• TransE
• TransH
• TransR
• TransD
• RotatE

Semantic Based

• RESCAL
• DistMult

Loss

• Pairwise Hinge Loss
• Pairwise Logistic Loss
• Binary Cross Entropy Loss
• Self Adversarial Negative Sampling Loss
• Square Error Loss

Score

• Dot
• Lp-Distance
• Squared Lp-Distance

Constraint

• Lp-Regularization
• Clip Constraint
• Nomalized Embedding
• Soft Constraint

Negative Sampling Strategy

• Uniform Strategy
• Typed Strategy