Research map / Federated learning and privacy / Communication-efficient FL
SoteriaFL: A Unified Framework for Private Federated Learning with Communication Compression
Zhize Li, Haoyu Zhao, Boyue Li, Yuejie Chi
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 |
|---|---|---|---|
| random-k sparsification compression (k=d/20) instead of uncompressed LDP-SGD/LDP-SVRG | CDP-SGD, SoteriaFL-SGD, SoteriaFL-SVRG · a9a logistic regression with nonconvex regularizer, eps=1,5,10 | convergence vs communication rounds | Hurt |
| random-k sparsification compression instead of uncompressed baselines | CDP-SGD, SoteriaFL · a9a, eps=1,5,10 | utility/training loss vs communication bits | Improved |
| shifted compression (SoteriaFL) instead of direct compression (CDP-SGD) (ablation) | SoteriaFL-SGD · a9a logistic regression | utility and training loss | Improved |
| SVRG variance reduction instead of SGD inside SoteriaFL (ablation) | SoteriaFL · a9a logistic regression | utility | Improved |
| random-k sparsification compression instead of uncompressed baselines | CDP-SGD, SoteriaFL · MNIST shallow NN, eps=1,2,4,8,16 | test accuracy vs communication bits | Improved |
| compression instead of uncompressed baselines as eps increases (weaker privacy) | compressed algorithms · MNIST shallow NN | gap in convergence vs rounds | Mixed |
| shifted compression (SoteriaFL-SGD) instead of direct compression (CDP-SGD) (ablation) | SoteriaFL-SGD · MNIST shallow NN | utility and test accuracy | Improved |
| SoteriaFL-SVRG instead of CDP-SGD | SoteriaFL-SVRG · MNIST shallow NN | utility/test accuracy | Hurt |
| shifted compression (SoteriaFL) with local-DP noise instead of CDP-SGD | LDP-FL theory · nonconvex LDP, theory | number of communication rounds for same utility | Improved |
Closest papers by meaning
- Achieving Dimension-Free Communication in Federated Learning via Zeroth-Order Optimization (ICLR 2025 · similarity 0.89)
- Privacy-Preserving Federated Convex Optimization: Balancing Partial-Participation and Efficiency via Noise Cancellation (ICML 2025 · similarity 0.88)
- z-SignFedAvg: A Unified Stochastic Sign-Based Compression for Federated Learning (AAAI 2024 · similarity 0.88)
- Clients Collaborate: Flexible Differentially Private Federated Learning with Guaranteed Improvement of Utility-Privacy Trade-off (ICML 2025 · similarity 0.88)
- FedLMT: Tackling System Heterogeneity of Federated Learning via Low-Rank Model Training with Theoretical Guarantees (ICML 2024 · similarity 0.88)
- Sketched Gaussian Mechanism for Private Federated Learning (NeurIPS 2025 · similarity 0.88)
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.
