Research map / Federated learning and privacy / Communication-efficient FL
FedAvg with Fine Tuning: Local Updates Lead to Representation Learning
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 |
|---|---|---|---|
| use FedAvg with multiple local updates (tau>=2) instead of D-GD (tau=1) (ablation) | FedAvg multi-task linear regression · Multi-task linear regression (d=100,k=5,M=40) | principal angle distance to ground-truth representation | Helped |
| use FedAvg (tau=2) instead of D-GD before fine-tuning on new client (ablation) | linear regression · Multi-task linear regression, new client | fine-tuned error | Helped |
| use FedAvg with tau=50 instead of D-SGD (ablation) | CNN on CIFAR-100 · Fine-tuning on 20 unseen CIFAR-100 classes | fine-tuned accuracy | Helped |
| use FedAvg with tau=50 instead of D-SGD (ablation) | CNN on CIFAR-100 · Fine-tuning on CIFAR-10 clients | fine-tuned accuracy | Helped |
| increase number of local updates tau (ablation) | FedAvg CNN on CIFAR-10 · Heterogeneous CIFAR-10 (Dirichlet 0.6, M=100) | class-wise feature cosine similarity | Helped |
| use D-GD instead of FedAvg (ablation) | linear regression · Multi-task linear regression | global loss gradient norm | Helped |
Closest papers by meaning
- Widening the Network Mitigates the Impact of Data Heterogeneity on FedAvg (ICML 2025 · similarity 0.91)
- Momentum Benefits Non-iid Federated Learning Simply and Provably (ICLR 2024 · similarity 0.89)
- FedImpro: Measuring and Improving Client Update in Federated Learning (ICLR 2024 · similarity 0.89)
- A Lightweight Method for Tackling Unknown Participation Statistics in Federated Averaging (ICLR 2024 · similarity 0.87)
- Exact and Linear Convergence for Federated Learning under Arbitrary Client Participation is Attainable (NeurIPS 2025 · similarity 0.87)
- FedLoGe: Joint Local and Generic Federated Learning under Long-tailed Data (ICLR 2024 · 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.