Embedding-reparameterization procedure for manifold-valued latent variables in generative models.
Eugene A. GolikovMaksim KretovPublished in: CoRR (2018)
Keyphrases
- latent variables
- generative model
- probabilistic model
- latent space
- topic models
- prior knowledge
- hidden variables
- posterior distribution
- latent structure
- latent variable models
- graphical models
- mixture model
- bayesian framework
- gaussian process
- approximate inference
- discriminative learning
- conditional random fields
- em algorithm
- language model
- latent dirichlet allocation
- random variables
- expectation maximization
- bayesian networks
- object categories
- structured prediction
- vector space
- decision trees
- sparse representation
- supervised learning
- multidimensional scaling
- high dimensional
- support vector
- image segmentation