mlsanityData to 14 Oct 2026

Private Matrix Factorization with Public Item Features

arXiv 2309.11516 ↗First posted 17 Sep 2023cs.IR cs.CR cs.LGcited by 3

Not on the map. Why?

Abstract (arXiv)

We consider the problem of training private recommendation models with access to public item features. Training with Differential Privacy (DP) offers strong privacy guarantees, at the expense of loss in recommendation quality. We show that incorporating public item features during training can help mitigate this loss in quality. We propose a general approach based on collective matrix factorization (CMF), that works by simultaneously factorizing two matrices: the user feedback matrix (representing sensitive data) and an item feature matrix that encodes publicly available (non-sensitive) item information. The method is conceptually simple, easy to tune, and highly scalable. It can be applied to different types of public item data, including: (1) categorical item features; (2) item-item similarities learned from public sources; and (3) publicly available user feedback. Furthermore, these data modalities can be collectively utilized to fully leverage public data. Evaluating our method on a standard DP recommendation benchmark, we find that using public item features significantly narrows the quality gap between private models and their non-private counterparts. As privacy constraints become more stringent, models rely more heavily on public side features for recommendation. This results in a smooth transition from collaborative filtering to item-based contextual recommendations.

Key facts. Cited by 3 papers in the map (data through 14 Oct 2026). Builds on 23 of its 54 references that are in the map. First posted on arXiv 17 Sep 2023.

Builds on

23 of 54 references.

References · 23

  1. 2016Deep Learning with Differential Privacycited by 2,728
  2. 2021Large Language Models Can Be Strong Differentially Private Learnerscited by 412
  3. 2021Differentially Private Fine-tuning of Language Modelscited by 400
  4. 2018Variational Autoencoders for Collaborative Filteringcited by 310
  5. 2020Bypassing the Ambient Dimension: Private SGD with Gradient Subspace Identificationcited by 90
  6. 2021Do Not Let Privacy Overbill Utility: Gradient Embedding Perturbation for Private Learningcited by 86
  7. 2018Revisiting differentially private linear regression: optimal and adaptive prediction & estimation in unbounded domaincited by 61
  8. 2021Membership Inference Attacks Against Recommender Systemscited by 57
  9. 2022Mixed Differential Privacy in Computer Visioncited by 54
  10. 2021Public Data-Assisted Mirror Descent for Private Model Trainingcited by 49
  11. 2017Unsupervised Learning of Semantic Audio Representationscited by 44
  12. 2022Privately Fine-Tuning Large Language Models with Differential Privacycited by 39
11 more
  1. 2022Private Adaptive Optimization with Side Informationcited by 38
  2. 2016On the Theory and Practice of Privacy-Preserving Bayesian Data Analysiscited by 35
  3. 2013Provable Inductive Matrix Completioncited by 32
  4. 2015Metadata Embeddings for User and Item Cold-start Recommendationscited by 29
  5. 2023Advancing Differential Privacy: Where We Are Now and Future Directions for Real-World Deploymentcited by 27
  6. 2022Learning to Generate Image Embeddings with User-level Differential Privacycited by 23
  7. 2020Learning from Mixtures of Private and Public Populationscited by 19
  8. 2021Private Alternating Least Squares: Practical Private Matrix Completion with Tighter Ratescited by 10
  9. 2017Differentially Private Matrix Completion Revisitedcited by 8
  10. 2018Cold-start recommendations in Collective Matrix Factorizationcited by 8
  11. 2023Multi-Task Differential Privacy Under Distribution Skewcited by 3

Built on by

Cited by year

Citing papers per year: 2023: 1; 2024: 1; 2025: 1.

3 papers.

  1. 2023Private Learning with Public Featurescited by 6
  2. 2024Locally Private Estimation with Public Featurescited by 3
  3. 2025Privacy-Utility-Bias Trade-offs for Privacy-Preserving Recommender Systemscited by 2