A backward pass through a CNN using a generative model of its activations.
Hua-Yan WangAnna ChenYi LiuDileep GeorgeD. Scott PhoenixPublished in: CoRR (2016)
Keyphrases
- generative model
- cellular neural networks
- probabilistic model
- em algorithm
- bayesian framework
- mixture model
- discriminative models
- prior knowledge
- posterior probability
- discriminative learning
- semi supervised
- object categories
- topic models
- generative and discriminative models
- latent dirichlet allocation
- computer vision
- markov chain monte carlo
- learned models
- low level
- feature vectors
- spatio temporal
- fully bayesian
- similarity measure