Difformer: Empowering Diffusion Model on Embedding Space for Text Generation.
Zhujin GaoJunliang GuoXu TanYongxin ZhuFang ZhangJiang BianLinli XuPublished in: CoRR (2022)
Keyphrases
- diffusion model
- text generation
- embedding space
- euclidean space
- natural language generation
- low dimensional
- manifold learning
- dimensionality reduction
- high dimensional
- graph embedding
- input space
- geometric structure
- natural language
- data points
- diffusion process
- information diffusion
- geodesic distance
- anisotropic diffusion
- theorem prover
- nonlinear dimensionality reduction
- semi supervised
- euclidean distance
- metric space
- feature vectors
- feature space
- vector space
- high dimensional data
- edge detection
- information extraction
- moving objects
- training set