Fast Differentially Private Matrix Factorization

2015/05/06 by Ziqi Liu, Yu-Xiang Wang, Liu, Ziqi +3 · 1 citation
Computer Science · Mathematics · #Artificial Intelligence (cs.AI) #FOS: Computer and information sciences #G.1.6 #G.2 #G.3 #I.2.6 #Machine Learning (cs.LG) #Privacy-Preserving Technologies in Data #Random Matrices and Applications #Stochastic Gradient Optimization Techniques

paper · pdf · doi:10.48550/arxiv.1505.01419

openalex publication_date 2015/05/06 · openalex created_date 2025/10/10 · openalex updated_date 2026/08/01

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.

Citations

• Guess Who Rated This Movie: Identifying Users Through Subspace Clustering
• Matrix Factorization Techniques for Recommender Systems
• GraphLab: A New Framework For Parallel Machine Learning
• HOGWILD!: A Lock-Free Approach to Parallelizing Stochastic Gradient Descent

Cited by

• Privacy-Utility-Bias Trade-offs for Privacy-Preserving Recommender Systems

Related

• Differentially Private Matrix Completion Revisited
• The Price of Differential Privacy for Low-Rank Factorization
• Multi-Epoch Matrix Factorization Mechanisms for Private Machine Learning
• (Amplified) Banded Matrix Factorization: A unified approach to private training
• Private Alternating Least Squares: Practical Private Matrix Completion with Tighter Rates