Privacy Amplification with Couplings and Overlapping Mixtures

Borja Balle, Amazon Research

When:
Thursday, June 20, 2019
Event Start Time: 1:30 pm
Event End Time: 3:00 pm
Where:
Hariri Institute for Computing, 111 Cummington Mall, Seminar Room, MCS 180

Abstract: Privacy amplification is a key component of many state-of-the-art analyses in differential privacy. Amplification results come under different guises, including amplification by sampling, shuffling and iteration. In this talk I will survey a number of recent results in this area, and try to shed some light on the common principles behind them.

Bio: From April 2017 until May 2019 Borja was senior machine learning scientist in Neil Lawrence’s team at Amazon (Cambridge, UK). Before that, he was a lecturer (assistant professor) at Lancaster University in the Department of Mathematics and Statistics as well as the Data Science Institute. From October 2013 until September 2015 he was a post-doctoral fellow in the Reasoning and Learning Laboratory at McGill University, where he worked with Prakash Panangaden, Joelle Pineau, and Doina Precup. He earned his PhD in 2013 from UPC after working with the LARCA research group under the supervision of Jorge Castro and Ricard Gavaldà. During his PhD he spent several months visiting Mehryar Mohriat at the Courant Institute (NYU).