A Universal Marginalizer for Amortized Inference in Generative Models.
Laura DouglasIliyan ZarovKonstantinos GourgouliasChris LucasChris HartAdam BakerManeesh SahaniYura PerovSaurabh JohriPublished in: CoRR (2017)
Keyphrases
- generative model
- boltzmann machine
- parameter learning
- probabilistic model
- discriminative learning
- representational power
- discriminative models
- conditional random fields
- mixture model
- bayesian framework
- em algorithm
- posterior probability
- bayesian networks
- pitman yor process
- hierarchical hidden markov models
- probabilistic inference
- structured prediction
- latent dirichlet allocation
- search tree
- prior knowledge
- semi supervised
- expectation maximization
- maximum entropy principle
- object categories
- generative and discriminative models
- generative process
- hidden variables
- bayesian inference
- natural images
- efficient learning
- belief networks
- computer vision