Orthogonality and graph divergence losses promote disentanglement in generative models.
Ankita ShuklaRishi DadhichRajhans SinghAnirudh RayasPouria SaidiGautam DasarathyVisar BerishaPavan K. TuragaPublished in: Frontiers Comput. Sci. (2024)
Keyphrases
- generative model
- probabilistic model
- mixture model
- discriminative models
- discriminative learning
- prior knowledge
- hierarchical hidden markov models
- semi supervised
- conditional random fields
- em algorithm
- hidden variables
- generative and discriminative models
- topic models
- maximum entropy principle
- spanning tree
- maximum likelihood
- expectation maximization
- graphical models
- object detection
- object categories
- pairwise
- representational power
- information retrieval