Research map / Federated learning and privacy / Personalized and heterogeneous FL
Client-level Active Error Correction in Distributed Learning
Kwang In 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 |
|---|---|---|---|
| use GPAEC gradient-GP client selection instead of random selection | client-level AEC · CIFAR100 / C1M / EMNIST (stage 11) | accuracy: C1M 53.99 -> 55.85; EMNIST 81.15 -> 82.42; CIFAR100 59.34 -> 60.51 | Improved |
| select clients by highest average loss instead of GPAEC | client-level AEC baseline · C1M early stage 1 | accuracy: Rand 49.08 -> Loss 48.70 (GPAEC 51.19) | Hurt |
| rank-combine GP inconsistency with loss deviation instead of loss alone | GPAEC · first stage client selection | fraction rational clients selected: 42.2% (Loss) -> 88.9% | Improved |
| replace GP inconsistency with distance from mean gradient (ablation) | GPAEC · CIFAR10 | accuracy at stage 11: 85.64 -> 84.47 | Hurt |
| replace GP inconsistency with distance from median gradient (ablation) | GPAEC · CIFAR10 | accuracy at stage 11: 85.64 -> 85.08 | Hurt |
| use small 3-conv CNN instead of ResNet50 (ablation) | GPAEC · Kuzushiji49 | accuracy at stage 11: GPAEC 47.81 vs Rand 46.43 | Improved |
| select L=60 clients per stage (ablation) | GPAEC · CIFAR100 | accuracy at stage 11: Rand 61.35 -> GPAEC 61.81 | Improved |
| select L=20 clients per stage (ablation) | GPAEC · TImgNet | accuracy at stage 11: Rand 50.83 -> GPAEC 52.64 | Improved |
| unreliable clients clean labels but inflate loss (ablation) | GPAEC · C1M | accuracy at stage 11: Rand 54.77 -> GPAEC 56.33 | Improved |
| select by entropy instead of GPAEC | client-level AEC baseline · C1M | accuracy at stage 11: Rand 53.99 -> Ent 52.58 | Hurt |
| spread budget uniformly over all clients instead of selecting | client-level AEC baseline · CIFAR10 | accuracy at stage 11: Rand 84.74 -> Uniform 85.02 | No clear effect |
Closest papers by meaning
- Collaborative Learning via Prediction Consensus (NeurIPS 2023 · similarity 0.85)
- Diffusion-Based Active Learning for Distributed Client Manifolds (AAAI 2025 · similarity 0.84)
- Unextractable Protocol Models: Collaborative Training and Inference without Weight Materialization (NeurIPS 2025 · similarity 0.83)
- Aequa: Fair Model Rewards in Collaborative Learning via Slimmable Networks (ICML 2025 · similarity 0.83)
- Enhancing Robustness of Last Layer Two-Stage Fair Model Corrections (NeurIPS 2024 · similarity 0.83)
- Coupled Confusion Correction: Learning from Crowds with Sparse Annotations (AAAI 2024 · similarity 0.82)
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.
