Hyper-Representations as Generative Models: Sampling Unseen Neural Network Weights.
Konstantin SchürholtBoris KnyazevXavier Giró-i-NietoDamian BorthPublished in: NeurIPS (2022)
Keyphrases
- generative model
- neural network
- feature representations
- probabilistic model
- mixture model
- discriminative learning
- discriminative models
- visual patterns
- generative and discriminative models
- conditional random fields
- deep belief networks
- hierarchical models
- hierarchical hidden markov models
- prior knowledge
- maximum entropy principle
- em algorithm
- posterior probability
- semi supervised
- training set
- linear combination
- artificial neural networks
- information retrieval
- object categories
- representational power
- learning rules
- topic models
- expectation maximization
- image classification
- active learning
- training data
- feature selection
- machine learning