Timezone: »

 
Poster
Differentially Private Markov Chain Monte Carlo
Mikko Heikkilä · Joonas Jälkö · Onur Dikmen · Antti Honkela

Thu Dec 12 05:00 PM -- 07:00 PM (PST) @ East Exhibition Hall B + C #158
in Applications -- Privacy, Anonymity, and Security »

Recent developments in differentially private (DP) machine learning and DP Bayesian learning have enabled learning under strong privacy guarantees for the training data subjects. In this paper, we further extend the applicability of DP Bayesian learning by presenting the first general DP Markov chain Monte Carlo (MCMC) algorithm whose privacy-guarantees are not subject to unrealistic assumptions on Markov chain convergence and that is applicable to posterior inference in arbitrary models. Our algorithm is based on a decomposition of the Barker acceptance test that allows evaluating the Rényi DP privacy cost of the accept-reject choice. We further show how to improve the DP guarantee through data subsampling and approximate acceptance tests.

Keywords: MCMC · Privacy, Anonymity, and Security · Applications · Probabilistic Methods
[ Paper]  [ Poster] 

Author Information

Mikko Heikkilä (University of Helsinki)
Joonas Jälkö (Aalto University)
Onur Dikmen (Halmstad University)
Antti Honkela (University of Helsinki)

Related Events (a corresponding poster, oral, or spotlight)

More from the Same Authors