How Does Fine-tuning Affect the Geometry of Embedding Space: A Case Study on Isotropy.
Sara RajaeeMohammad Taher PilehvarPublished in: EMNLP (Findings) (2021)
Keyphrases
- fine tuning
- embedding space
- geometric structure
- geodesic distance
- manifold learning
- graph embedding
- low dimensional
- euclidean space
- high dimensional
- dimensionality reduction
- fine tuned
- input space
- data points
- discriminant analysis
- relative position
- shape analysis
- dimension reduction
- subspace learning
- kernel methods
- nonlinear dimensionality reduction
- nearest neighbor
- data sets