Research map / Federated learning and privacy / Communication-efficient FL
FedQueue: Queue-Aware Federated Learning for Cross-Facility HPC Training
Yijiang Li, Emon Dey, Zilinghan Li, Krishnan Raghavan, Ravi K Madduri, Kibaek Kim
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 |
|---|---|---|---|
| add EWMA queue-delay prediction (ablation) | FedQueue · Synthetic non-IID MNIST | time to 85% accuracy: 116.59 -> 26.68 | Improved |
| add EWMA queue-delay prediction (ablation) | FedQueue · Synthetic non-IID MNIST | final accuracy: 96.01 -> 96.62 | No clear effect |
| add staleness decay in aggregation (ablation) | FedQueue · Synthetic non-IID MNIST | final accuracy: 87.18 -> 96.62 | Improved |
| add inverse learning-rate scaling (ablation) | FedQueue · Synthetic non-IID MNIST | final accuracy: 90.89 -> 96.62 | Improved |
| increase safety buffer delta (ablation) | FedQueue · Synthetic queues | time to 95% accuracy and staleness | Mixed |
| use EWMA rate alpha 0.5 instead of 0.1 or 1.0 (ablation) | FedQueue · Synthetic queues | late arrivals | Improved |
| use queue-aware FedQueue instead of FedCompass/FedAvg | federated learning · Real HPC deployment, LLaMA2-7B | final test loss: 0.4345 -> 0.3947 | Improved |
| use FedQueue instead of baselines under high queue variance | federated learning · Synthetic non-IID MNIST, rho=0.9 | time to 95% accuracy: 245.60 -> 154.72 (vs FedAvg) | Improved |
| use FedQueue instead of FedBuff/FedCompass | federated learning · Synthetic non-IID MNIST, rho=0.9 | max accuracy: 97.45 -> 96.62 (vs FedBuff) | Hurt |
| use FedQueue instead of FedBuff/FedCompass | federated learning · CIFAR-100 IID | max accuracy: 56.4 -> 51.7 (vs FedBuff) | Hurt |
| use FedQueue instead of baselines | federated learning · CIFAR-100 non-IID | max accuracy: 44.7 -> 46.9 (vs FedCompass) | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- FedCompass — FedCompass: Efficient Cross-Silo Federated Learning on Heterogeneous Client Devices Using a Computing Power-Aware Scheduler (ICLR 2024) · better on 5 of 6 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- FedCompass: Efficient Cross-Silo Federated Learning on Heterogeneous Client Devices Using a Computing Power-Aware Scheduler (ICLR 2024 · similarity 0.86)
- FedASMU: Efficient Asynchronous Federated Learning with Dynamic Staleness-Aware Model Update (AAAI 2024 · similarity 0.85)
- Reviving Stale Updates: Data-Free Knowledge Distillation for Asynchronous Federated Learning (NeurIPS 2026 · similarity 0.85)
- HERO: A Heterogeneity-Aware Benchmark Library for Federated Continual Learning (NeurIPS 2026 · similarity 0.84)
- Towards Straggler-Resilient Split Federated Learning: An Unbalanced Update Approach (NeurIPS 2025 · similarity 0.83)
- Efficient Asynchronous Federated Learning with Prospective Momentum Aggregation and Fine-Grained Correction (AAAI 2024 · similarity 0.83)
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.
