Research map / Federated learning and privacy / Differential privacy
Differentially Private Matrix Completion Revisited
Prateek Jain, Om Dipakbhai Thakkar, Abhradeep Thakurta
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use Private FW rather than prior private matrix-completion algorithms | User-level joint-DP matrix completion · Jester, MovieLens10M, Netflix, Yahoo! Music, and synthetic data | test RMSE | Improved |
| Replace dense private covariance eigensolving with Private Oja's method (ablation) | Private FW · Collaborative-filtering experiments | test RMSE | No clear effect |
| Use rank-one Frank-Wolfe updates instead of private PGD updates | Private matrix completion · Collaborative-filtering experiments | test RMSE | Improved |
| Use Private SVD instead of Private FW | Private matrix completion · Settings satisfying the SVD method's incoherence assumptions | generalization error | Mixed |
| Increase the number of users | Private FW · Matrix completion | generalization error | Improved |
| Require more users than non-private nuclear norm minimization | Private FW · Matrix completion sample complexity | required users m: $\omega(n)$ to $\omega(n^{5/4})$ | Hurt |
Closest papers by meaning
- Private Alternating Least Squares: Practical Private Matrix Completion with Tighter Rates (ICML 2021 · similarity 0.90)
- The Price of Privacy for Low-rank Factorization (NeurIPS 2018 · similarity 0.89)
- A Framework for Private Matrix Analysis in Sliding Window Model (ICML 2021 · similarity 0.88)
- Improved Differential Privacy for SGD via Optimal Private Linear Operators on Adaptive Streams (NeurIPS 2022 · similarity 0.87)
- Sampling-Free Privacy Accounting for Matrix Mechanisms under Random Allocation (NeurIPS 2026 · similarity 0.87)
- Differentially Private Submodular Maximization with a Knapsack Constraint (ICML 2026 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
