Research map / Optimization / Distributed and compressed
Improving the Privacy and Accuracy of ADMM-Based Distributed Algorithms
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Increase the penalty parameter geometrically over iterations | Private M-ADMM with penalty perturbation (PP) · Adult | training average loss | Helped |
| Increase the penalty parameter geometrically over iterations | Private M-ADMM with penalty perturbation (PP) · Adult | privacy loss upper bound | Helped |
| Increase the penalty parameter | M-ADMM · Adult, five-node network | convergence speed | Hurt |
| Use penalty perturbation instead of dual variable perturbation | Private M-ADMM · Adult | training average loss | Mixed |
| Use penalty perturbation instead of dual variable perturbation | Private M-ADMM · Adult | privacy loss upper bound | Mixed |
| Increase the noise level | Private M-ADMM · Adult | training average loss | Hurt |
| Increase the noise level | Private M-ADMM · Adult | privacy loss upper bound | Helped |
| Keep each node's penalty parameter private | Private M-ADMM · Adult | privacy protection against inference | Helped |
| Use the same fixed penalty and dual-update step size at every node and iteration (ablation) | Private M-ADMM (dual variable perturbation special case) · Adult | privacy-utility tradeoff | Hurt |
| Increase the penalty parameter over iterations | M-ADMM · Adult, larger networks | convergence | No clear effect |
| Use uneven rather than even sample allocation across nodes (ablation) | Private M-ADMM · Adult, 100-node network | privacy loss | Hurt |
Closest papers by meaning
- From Noisy Fixed-Point Iterations to Private ADMM for Centralized and Federated Learning (ICML 2023 · similarity 0.92)
- Privacy Amplification by Iteration for ADMM with (Strongly) Convex Objective Functions (AAAI 2024 · similarity 0.91)
- Distributed Learning without Distress: Privacy-Preserving Empirical Risk Minimization (NeurIPS 2018 · similarity 0.87)
- Noisy SIGNSGD Is More Differentially Private Than You (Might) Think (ICML 2025 · similarity 0.85)
- Revisiting Differentially Private Algorithms for Decentralized Online Learning (ICML 2025 · similarity 0.85)
- Differentially Private Bilevel Optimization: Efficient Algorithms with Near-Optimal Rates (NeurIPS 2025 · similarity 0.85)
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.