Research map / Multi-agent systems and games / Games and equilibria
Byzantine Resilient Distributed Multi-Task Learning
Jiani Li, Waseem Abbas, Xenofon Koutsoukos
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 loss-based neighbor weights with filtering instead of non-cooperative SGD | Distributed multi-task learning · Target Localization | average loss | Improved |
| Use loss-based neighbor weights with filtering instead of non-cooperative SGD | Distributed multi-task learning · Human Activity Recognition | average testing loss and accuracy | Improved |
| Use loss-based neighbor weights with filtering instead of non-cooperative SGD | Distributed multi-task learning · Digit Classification, MNIST and synthetic digits | average testing loss and accuracy | Improved |
| Use loss-based weights when every neighbor is Byzantine, rather than standard non-cooperative SGD | A normal agent in distributed multi-task learning · Target Localization, Human Activity Recognition, and Digit Classification | average loss and accuracy | No clear effect |
| Measure similarity by accumulated loss on the receiving agent's data rather than model-parameter distance | Online neighbor weighting · Byzantine attacks | resilience to malicious model updates | Improved |
| Filter neighbors whose loss exceeds the receiving agent's own loss | Loss-based neighbor weighting · Byzantine-resilient convergence | expected regret | Improved |
| Cooperate with neighbors using loss-based weights rather than learn alone | A normal agent in distributed multi-task learning · Strongly convex risk functions, arbitrary Byzantine neighbors | expected regret bound: \mathbb{E} [r_k(\theta_{k,i}) - r_k(\theta_k^*)] \leq \mathbb{E} [r_k(\hat{\theta}_{k,i}) - r_k(\theta_k^*)] | Improved |
Closest papers by meaning
- ResMAS: Resilience Optimization in LLM-based Multi-agent Systems (AAAI 2026 · similarity 0.86)
- Byzantine-Robust Decentralized Learning via Remove-then-Clip Aggregation (AAAI 2024 · similarity 0.86)
- From Inexact Gradients to Byzantine Robustness: Acceleration and Optimization under Similarity (NeurIPS 2026 · similarity 0.85)
- Weight for Robustness: A Comprehensive Approach towards Optimal Fault-Tolerant Asynchronous ML (NeurIPS 2024 · similarity 0.85)
- Fault Tolerant ML: Efficient Meta-Aggregation and Synchronous Training (ICML 2024 · similarity 0.85)
- On the Tension between Byzantine Robustness and No-Attack Accuracy in Distributed Learning (ICML 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.
