Research map / Federated learning and privacy / Communication-efficient FL
Workie-Talkie: Accelerating Federated Learning by Overlapping Computing and Communications via Contrastive Regularization
Rui Chen, Qiyu Wan, Pavana Prakash, Lan Zhang, Xu Yuan, Yanmin Gong, Xin Fu, Miao Pan
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 class-wise temperature instead of fixed temperature in contrastive regularization (ablation) | FedCR · CIFAR-100 (beta=0.5) | test accuracy: 61.3% -> 62.3% | Improved |
| use class-wise temperature instead of average temperature (ablation) | FedCR · CIFAR-100 (beta=0.5) | test accuracy: 61.9% -> 62.3% | Improved |
| class-wise temperature gain grows as data heterogeneity increases (beta smaller) (ablation) | FedCR · CIFAR-10/CIFAR-100 | test accuracy: CIFAR-10 beta=0.5: 82.2% -> 83.1%; beta=5: 87.2% -> 88.2% | Improved |
| contrastive regularization with fixed projector instead of gradient correction (DGA) | overlapped FL · CIFAR-100 | test accuracy: 58.8% -> 62.3% | Improved |
| replace trainable projector (Overlap-MOON) with fixed non-trainable projector | FedCR · CIFAR-100 | test accuracy / speedup: 63.6% -> 62.3% accuracy; speedup 2.18 -> 2.58 | Mixed |
| overlap computing and communication with FedCR vs DGA/LGC | FL training · CIFAR-100 target 58.8% accuracy | training time speedup: 1.52x vs DGA, 2.01x vs LGC | Improved |
| FedCR vs DGA/LGC rounds on Tiny-ImageNet | FL training · Tiny-ImageNet target 20% | communication rounds: 1.65x and 2.67x fewer rounds vs DGA and LGC | Improved |
| increase staleness level (DGA gradient correction) (ablation) | DGA · CIFAR-10 ResNet20 CKA | CKA similarity local vs global: 0.72 -> 0.51 | Hurt |
| train heterogeneous subnetworks with FedCR instead of Overlap-HeteroFL | FedCR · CIFAR-100 p-subnetworks | test accuracy: +23%, +15%, +37% on different p-subnetworks | Improved |
| FedCR vs Overlap-MOON with more devices and time budget | FedCR · CIFAR-100 N=64, 25x10^2 s | test accuracy: 58.01% -> 64.68% | Improved |
Closest papers by meaning
- WHALE-FL: Wireless and Heterogeneity Aware Latency Efficient Federated Learning over Mobile Devices via Adaptive Subnetwork Scheduling (AAAI 2025 · similarity 0.91)
- Low Precision Local Training is Enough for Federated Learning (NeurIPS 2024 · similarity 0.87)
- FedLPS: Heterogeneous Federated Learning for Multiple Tasks with Local Parameter Sharing (AAAI 2024 · similarity 0.87)
- FedCompass: Efficient Cross-Silo Federated Learning on Heterogeneous Client Devices Using a Computing Power-Aware Scheduler (ICLR 2024 · similarity 0.87)
- FedQS: Optimizing Gradient and Model Aggregation for Semi-Asynchronous Federated Learning (NeurIPS 2025 · similarity 0.87)
- SparsyFed: Sparse Adaptive Federated Learning (ICLR 2025 · similarity 0.87)
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.
