Generalization in Generative Adversarial Networks: A Novel Perspective from Privacy Protection
arXiv:1908.07882
Abstract
In this paper, we aim to understand the generalization properties of generative adversarial networks (GANs) from a new perspective of privacy protection. Theoretically, we prove that a differentially private learning algorithm used for training the GAN does not overfit to a certain degree, i.e., the generalization gap can be bounded. Moreover, some recent works, such as the Bayesian GAN, can be re-interpreted based on our theoretical insight from privacy protection. Quantitatively, to evaluate the information leakage of well-trained GAN models, we perform various membership attacks on these models. The results show that previous Lipschitz regularization techniques are effective in not only reducing the generalization gap but also alleviating the information leakage of the training dataset.
Accepted by NeurIPS 2019
References in corpus (6)
- Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks
- Generalization and Equilibrium in Generative Adversarial Nets (GANs)
- Loss-Sensitive Generative Adversarial Networks on Lipschitz Densities
- Privacy for Free: Posterior Sampling and Stochastic Gradient Monte Carlo
- Learning with Differential Privacy: Stability, Learnability and the Sufficiency and Necessity of ERM Principle
- SRPGAN: Perceptual Generative Adversarial Network for Single Image Super Resolution
Cited by in corpus (9)
- Generative Adversarial Networks: A Survey Towards Private and Secure Applications
- Synthetic Observational Health Data with GANs: from slow adoption to a boom in medical research and ultimately digital twins?
- Generative Convolution Layer for Image Generation
- Input Perturbation: A New Paradigm between Central and Local Differential Privacy
- Weighted Distributed Differential Privacy ERM: Convex and Non-convex
- Characterizing Membership Privacy in Stochastic Gradient Langevin Dynamics
- Alleviation of Gradient Exploding in GANs: Fake Can Be Real
- Anonymization of labeled TOF-MRA images for brain vessel segmentation using generative adversarial networks
- Improving the expressiveness of neural vocoding with non-affine Normalizing Flows