Differentially Private Sampling from Distributions
A Google TechTalk, presented by Marika Swanberg, 2023-08-22
Google Algorithms Seminar. ABSTRACT: We initiate an investigation of private sampling from distributions. Given a dataset with n independent observations from an unknown distribution P, a sampling algorithm must output a single observation from a distribution that is close in total variation distance to P while satisfying differential privacy. Sampling abstracts the goal of generating small amounts of realistic-looking data.
We provide tight upper and lower bounds for the dataset size needed for this task for three natural families of distributions: arbitrary distributions on {1,…,k}, arbitrary product distributions on {0,1}^d, and product distributions on {0,1}^d with bias in each coordinate bounded away from 0 and 1. We demonstrate that, in some parameter regimes, private sampling requires asymptotically fewer observations than learning a description of P nonprivately; in other regimes, however, private sampling proves to be as difficult as private learning. Notably, for some classes of distributions, the overhead in the number of observations needed for private learning compared to non-private learning is completely captured by the number of observations needed for private sampling.
This work appeared at NeurIPS and the full version can be found here:
Bio: Marika is a rising fifth year PhD student at Boston University advised by Adam Smith. Her research spans differential privacy, cryptography, and their intersection with legal questions, and more recently she has become interested in practical implementations of differential privacy. Before interning at Google, she was a visiting assistant professor at Reed College, and she has also done research for Tumult Labs.
1 view
0
0
2 months ago 00:10:47 1
Advanced Verbs (C1-C2) for Total English Fluency
3 months ago 00:13:39 1
IMPERATIVES – How to give commands in English
4 months ago 00:03:21 1
INVISIBLE INFLUENCE: The Hidden Forces that Shape Behavior by Jonah Berger
5 months ago 01:22:25 1
The CEO of PROTON answers YOUR questions! Drive, Linux support, Photos, features, and a lot more!
6 months ago 00:41:58 2
Elliott Carter - String Quartet No. 1
6 months ago 00:08:59 1
Virtual visit of the International Court of Justice (ICJ)
6 months ago 00:19:18 30
DEEP ASMR CRACKING BACK MASSAGE AND CHIROPRACTIC ADJUSTMENTS FOR BEAUTIFUL GIRL LISA
7 months ago 00:28:13 1
PART IV: 20 ADVANCED VERBS to be more CONCISE in English! It’s not always about speaking faster...
7 months ago 00:02:15 5
Allan Rayman - Graceland | choreo by Lera Kazanina | stripdance стрип frame up танец на стуле приват
8 months ago 00:01:50 1
Ferrari 308 GrB & Gr4 | Pure V8 Sound ! Tour Auto 2021
9 months ago 00:04:05 2
Double Differential Steering System on a MODELCAR
9 months ago 00:44:49 1
Differentially Private Sampling from Distributions
9 months ago 00:08:36 1
All-New 2024 Mercedes Benz V-class - Best Luxury Van Interior Color Options
9 months ago 00:11:30 2
The U.S. Navy Just Admitted They’ve Created Something So Advanced & Said It Can’t Be Stopped
9 months ago 00:09:56 1
Rocco prepara la BRUSCHETTA TOSCANA a Fiesole tra le COLONNE ETRUSCHE + vista panoramica di Firenze!
10 months ago 00:08:52 1
Monk Solves His Latest Case Alongside Another Detective | Monk
11 months ago 00:00:00 1
Oh Twitch...... - WAN Show June 16, 2023
12 months ago 00:05:12 2
Marilyn Monroe (пародия) -Театр Пародий Анатолия Евдокимова Evdokimov Show/Лучшее шоу страны
12 months ago 00:02:11 1
Design Differential Op Amp Calculator Visual Basic P7
12 months ago 01:42:15 1
Yemen, Egypt or Madinah? - Ustaadh Abu Taymiyyah #InTheMaktaba
1 year ago 01:13:08 1
Insights 2015: K-HOLE, New York
1 year ago 00:16:34 1
Mark Reizen; “Wotan’s Farewell“; DIE WALKÜRE; Richard Wagner
1 year ago 00:04:12 1
Orlando Bloom Was Robbed by The Bling Ring
1 year ago 00:06:25 1
Class 12th – Theorem on Differentiability | Continuity and Differentiability | Tutorials Point