Reparameterized Sampling for Generative Adversarial Networks
arXiv:2107.00352
Abstract
Recently, sampling methods have been successfully applied to enhance the sample quality of Generative Adversarial Networks (GANs). However, in practice, they typically have poor sample efficiency because of the independent proposal sampling from the generator. In this work, we propose REP-GAN, a novel sampling method that allows general dependent proposals by REParameterizing the Markov chains into the latent space of the generator. Theoretically, we show that our reparameterized proposal admits a closed-form Metropolis-Hastings acceptance ratio. Empirically, extensive experiments on synthetic and real datasets demonstrate that our REP-GAN largely improves the sample efficiency and obtains better sample quality simultaneously.
ECML PKDD 2021
References in corpus (5)
- Markov Chain Monte Carlo and Variational Inference: Bridging the Gap
- NeuTra-lizing Bad Geometry in Hamiltonian Monte Carlo Using Neural Transport
- Normalizing Flows on Riemannian Manifolds
- Learning Model Reparametrizations: Implicit Variational Inference by Fitting MCMC distributions
- Discriminator Contrastive Divergence: Semi-Amortized Generative Modeling by Exploring Energy of the Discriminator