KGEmbeddingType
Enums · AiDotNet.Enums
Specifies the type of knowledge graph embedding model to use.
For Beginners
Each embedding type has strengths for different relationship patterns:
- TransE: Good for one-to-one relations (born_in, capital_of)
- RotatE: Handles symmetric, antisymmetric, inversion, and composition patterns
- ComplEx: Best for symmetric and antisymmetric relations
- DistMult: Best for symmetric relations (similar_to, married_to)
- TemporalTransE: TransE with time-awareness for facts that change over time
How It Works
Knowledge graph embeddings map entities and relations to continuous vector spaces, enabling mathematical reasoning about graph structure. Different models capture different types of relational patterns.
Fields
| Field | Summary |
|---|---|
KGEmbeddingType ComplEx |
ComplEx: Complex-valued embedding using Hermitian dot product. |
KGEmbeddingType DistMult |
DistMult: Bilinear diagonal model scoring Σ(h_k · r_k · t_k). |
KGEmbeddingType RotatE |
RotatE: Rotation-based embedding in complex space where t = h ∘ r. |
KGEmbeddingType TemporalTransE |
TemporalTransE: Time-aware TransE with discretized time bins. |
KGEmbeddingType TransE |
TransE: Translational embedding where h + r ≈ t. |
int value__ |