Hyperbolic Geometry is Not Necessary: Lightweight Euclidean-Based Models for Low-Dimensional Knowledge Graph Embeddings.
Kai WangYu LiuDan LinMichael ShengPublished in: EMNLP (Findings) (2021)
Keyphrases
- lightweight
- low dimensional
- euclidean space
- high dimensional
- domain knowledge
- prior knowledge
- vector space
- wireless sensor networks
- dimensionality reduction
- high dimensional data
- communication infrastructure
- data points
- principal component analysis
- manifold learning
- euclidean geometry
- knowledge base
- graph embedding
- probabilistic model
- feature space