Score-based generative models break the curse of dimensionality in learning a family of sub-Gaussian probability distributions.
Frank ColeYulong LuPublished in: CoRR (2024)
Keyphrases
- generative model
- discriminative learning
- learning process
- prior knowledge
- probability distribution
- learning algorithm
- parameter learning
- learning tasks
- hidden variables
- reinforcement learning
- em algorithm
- maximum entropy principle
- structured prediction
- conditional random fields
- maximum likelihood
- probabilistic model
- image segmentation