Abstract | ||
---|---|---|
The Probably Approximately Correct (PAC) Bayes framework (McAllester, 1999) can incorporate knowledge about the learning algorithm and (data) distribution through the use of distribution-dependent priors, yielding tighter generalization bounds on data-dependent posteriors. Using this flexibility, however, is difficult, especially when the data distribution is presumed to be unknown. We show how an e-differentially private data-dependent prior yields a valid PAC-Bayes bound, and then show how non-private mechanisms for choosing priors can also yield generalization bounds. As an application of this result, we show that a Gaussian prior mean chosen via stochastic gradient Langevin dynamics (SGLD; Welling and Teh, 2011) leads to a valid PAC-Bayes bound given control of the 2-Wasserstein distance to an epsilon-differentially private stationary distribution. We study our data-dependent bounds empirically, and show that they can be nonvacuous even when other distribution-dependent bounds are vacuous. |
Year | Venue | Keywords |
---|---|---|
2018 | ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 31 (NIPS 2018) | stochastic gradient langevin dynamics,learning algorithm,differential privacy,data distribution,probably approximately correct |
DocType | Volume | ISSN |
Conference | 31 | 1049-5258 |
Citations | PageRank | References |
4 | 0.37 | 9 |
Authors | ||
2 |
Name | Order | Citations | PageRank |
---|---|---|---|
Dziugaite, Gintare Karolina | 1 | 8 | 2.45 |
Daniel M. Roy | 2 | 818 | 63.27 |