Training Deep Generative Models in Highly Incomplete Data Scenarios With Prior Regularization.
Edgar A. BernalPublished in: CVPR Workshops (2021)
Keyphrases
- incomplete data
- generative model
- em algorithm
- deep belief networks
- prior knowledge
- hidden variables
- mixture model
- expectation maximization
- discriminatively trained
- restricted boltzmann machine
- probabilistic model
- missing data
- maximum a posteriori
- discriminative learning
- missing values
- bayesian framework
- prior information
- bayesian networks
- hyperparameters
- conditional random fields
- deep learning
- maximum likelihood
- discriminative models
- hierarchical hidden markov models
- supervised learning
- latent dirichlet allocation
- topic models
- training set
- generative and discriminative models
- unsupervised learning
- input data
- training data
- data mining