Self-supervised Disentanglement of Modality-Specific and Shared Factors Improves Multimodal Generative Models.
Imant DaunhawerThomas M. SutterRicards MarcinkevicsJulia E. VogtPublished in: GCPR (2020)
Keyphrases
- generative model
- multi modal
- discriminative learning
- probabilistic model
- mixture model
- discriminative models
- em algorithm
- prior knowledge
- conditional random fields
- object categories
- topic models
- hierarchical hidden markov models
- representational power
- expectation maximization
- semi supervised
- maximum entropy principle
- image processing
- hidden variables
- object categorization
- support vector
- naive bayes models
- generative and discriminative models
- decision trees