EE376b: Topics in Information Theory and Its Applications presents "Optimality in Locally Private Estimation"

Topic: 
Optimality in Locally Private Estimation
Thursday, May 16, 2019 - 10:00am
Venue: 
Packard 202
Speaker: 
John Duchi (Stanford)
Abstract / Description: 

I will present new optimality results for estimation problems in local privacy models, when data is kept private even from the collector of the data. I will give both fundamental limits—lower bounds—building out of communication complexity, allowing the limits to apply to any level of desired privacy and any mode of data collection. I will also give commensurate optimality results, showing in particular applications to large-scale estimation currently employed in a number of real-world scenarios.