Research map / Optimization / Distributed and compressed
Event-Triggered Gossip for Distributed Learning
Zhiyuan Zhai, Xiaojun Yuan, Xin Wang, Rui Zhang, Geoffrey Ye Li
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 event-triggered gossip (eps=0.005) instead of full communication | distributed SGD gossip · MNIST | accuracy: 0.9219 -> 0.9135 | No clear effect |
| use event-triggered gossip (eps=0.005) instead of full communication | distributed SGD gossip · Fashion-MNIST | accuracy: 0.7422 -> 0.7360 | No clear effect |
| cut transmissions with event triggering at eps=0.005 | event-triggered gossip · MNIST | cumulative communication: 39900 -> 12230 | Improved |
| cut transmissions with event triggering at eps=0.005 | event-triggered gossip · Fashion-MNIST | cumulative communication: 39900 -> 11328 | Improved |
| increase threshold from eps=5e-3 to 9e-3 (ablation) | event-triggered gossip · MNIST | accuracy: 0.9135 -> 0.8711 | Hurt |
| increase threshold from eps=5e-3 to 9e-3 (ablation) | event-triggered gossip · Fashion-MNIST | accuracy: 0.7360 -> 0.7292 | Hurt |
| increase threshold from eps=3e-3 to 7e-3 (communication) (ablation) | event-triggered gossip · MNIST | cumulative communication: 18934 -> 8893 | Improved |
| increase network sparsity (ablation) | event-triggered gossip · MNIST, Fashion-MNIST | accuracy | Mixed |
| use event-triggered instead of periodic gossip (Kp=5) | gossip scheme · MNIST | accuracy: 0.8454 -> 0.9135 | Improved |
| use event-triggered instead of variable working nodes | gossip scheme · MNIST | accuracy: 0.7406 -> 0.9135 | Improved |
| use event-triggered instead of probabilistic gossip | gossip scheme · MNIST | cumulative communication: 19904 -> 12230 | Improved |
| use event-triggered instead of periodic gossip (Kp=5) | gossip scheme · Fashion-MNIST | accuracy: 0.6884 -> 0.7360 | Improved |
Closest papers by meaning
- Distributed Event-Based Learning via ADMM (ICML 2025 · similarity 0.90)
- PDUDT: Provable Decentralized Unlearning under Dynamic Topologies (ICML 2025 · similarity 0.85)
- Learning Broadcast Protocols (AAAI 2024 · similarity 0.85)
- B-ary Tree Push-Pull Method is Provably Efficient for Distributed Learning on Heterogeneous Data (NeurIPS 2024 · similarity 0.84)
- Tight analyses of first-order methods with error feedback (NeurIPS 2025 · similarity 0.84)
- Unveiling the Power of Multiple Gossip Steps: A Stability-Based Generalization Analysis in Decentralized Training (NeurIPS 2025 · similarity 0.84)
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.
