Don't Generate Me: Training Differentially Private Generative Models with Sinkhorn Divergence.
Tianshi CaoAlex BieArash VahdatSanja FidlerKarsten KreisPublished in: NeurIPS (2021)
Keyphrases
- generative model
- differentially private
- probabilistic model
- discriminatively trained
- discriminative learning
- differential privacy
- prior knowledge
- discriminative models
- hierarchical hidden markov models
- generative and discriminative models
- em algorithm
- structured prediction
- supervised learning
- conditional random fields
- expectation maximization
- training examples
- data sets
- deep belief networks
- semi supervised
- data mining