Research map / Optimization / Distributed and compressed
Byzantine-Robust Distributed Learning: Towards Optimal Statistical Rates
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 |
|---|---|---|---|
| Replace mean aggregation with coordinate-wise median aggregation of worker gradients | distributed GD with 5% Byzantine machines · MNIST logistic regression, m=40 | test accuracy (%): 76.8 -> 87.2 | Helped |
| Replace mean aggregation with coordinate-wise trimmed mean (beta=0.05) of worker gradients | distributed GD with 5% Byzantine machines · MNIST logistic regression, m=40 | test accuracy (%): 76.8 -> 86.9 | Helped |
| Replace mean aggregation with coordinate-wise median aggregation | distributed GD with 10% Byzantine machines · MNIST CNN, m=10 | test accuracy (%): 77.3 -> 87.4 | Helped |
| Replace mean aggregation with coordinate-wise trimmed mean (beta=0.1) | distributed GD with 10% Byzantine machines · MNIST CNN, m=10 | test accuracy (%): 77.3 -> 90.7 | Helped |
| Replace mean aggregation of local ERMs with coordinate-wise median (one-round) | one-round distributed algorithm with 10% Byzantine machines · MNIST logistic regression, m=10 | test accuracy (%): 83.7 -> 89.0 | Helped |
| Trimmed-mean GD achieves order-optimal rate vs median GD which has extra 1/n term (theory comparison) | robust distributed GD · strongly convex losses, theory | error rate | No clear effect |
Closest papers by meaning
- From Inexact Gradients to Byzantine Robustness: Acceleration and Optimization under Similarity (NeurIPS 2026 · similarity 0.94)
- On the Tension between Byzantine Robustness and No-Attack Accuracy in Distributed Learning (ICML 2025 · similarity 0.91)
- Fault Tolerant ML: Efficient Meta-Aggregation and Synchronous Training (ICML 2024 · similarity 0.90)
- Byzantine Robustness and Partial Participation Can Be Achieved at Once: Just Clip Gradient Differences (NeurIPS 2024 · similarity 0.90)
- Weight for Robustness: A Comprehensive Approach towards Optimal Fault-Tolerant Asynchronous ML (NeurIPS 2024 · similarity 0.90)
- Byzantine-Robust Decentralized Learning via Remove-then-Clip Aggregation (AAAI 2024 · similarity 0.89)
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.