Fast Differentially Private Matrix Factorization
- 16 September 2015
- proceedings article
- conference paper
- Published by Association for Computing Machinery (ACM)
Abstract
Differentially private collaborative filtering is a challenging task, both in terms of accuracy and speed. We present a simple algorithm that is provably differentially private, while offering good performance, using a novel connection of differential privacy to Bayesian posterior sampling via Stochastic Gradient Langevin Dynamics. Due to its simplicity the algorithm lends itself to efficient implementation. By careful systems design and by exploiting the power law behavior of the data to maximize CPU cache bandwidth we are able to generate 1024 dimensional models at a rate of 8.5 million recommendations per second on a single PC.Keywords
Other Versions
Funding Information
- National Science Foundation (BCS-0941518)
- National Natural Science Foundation of China (61428206,91118005,91218301)
- Ministry of Education of the People's Republic of China (IRT13035)
- Science Fund for Creative Research Groups (61221063)
- National Research Foundation-Prime Minister's office, Republic of Singapore
This publication has 18 references indexed in Scilit:
- Large-Scale Distributed Bayesian Matrix Factorization using Stochastic Gradient MCMCPublished by Association for Computing Machinery (ACM) ,2015
- Differential PrivacyPublished by Association for Computing Machinery (ACM) ,2015
- Robust and Private Bayesian InferenceLecture Notes in Computer Science, 2014
- Scalable inference in latent variable modelsPublished by Association for Computing Machinery (ACM) ,2012
- MCMC Using Hamiltonian DynamicsPublished by Informa UK Limited ,2011
- Robust De-anonymization of Large Sparse DatasetsPublished by Institute of Electrical and Electronics Engineers (IEEE) ,2008
- Toward trustworthy recommender systemsACM Transactions on Internet Technology, 2007
- Differential PrivacyLecture Notes in Computer Science, 2006
- Calibrating Noise to Sensitivity in Private Data AnalysisLecture Notes in Computer Science, 2006
- Fast Generation of Discrete Random VariablesJournal of Statistical Software, 2004