Research map / Federated learning and privacy / Personalized and heterogeneous FL
Benchmarking Federated Learning & Knowledge Distillation for Point Cloud Classification
Aizierjiang Aiersilan
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 |
|---|---|---|---|
| train with federated label-skew non-IID instead of centralized | PointNet++ SSG (best FL method FedNova) · ModelNet40 | instance accuracy: 92.26 -> 76.32 | Hurt |
| train with federated label-skew non-IID instead of centralized | PointNet++ SSG (best FL method FedProx) · Craniosynostosis | instance accuracy: 100.00 -> 75.83 | Hurt |
| use server-side optimizers (FedAvgM, FedAdam, FedYogi, FedAdagrad) instead of FedAvg | PointNet++ SSG · ModelNet40 | instance accuracy: 58.51 -> 4.05 | Hurt |
| use FedProx proximal term instead of FedAvg | PointNet++ SSG · ModelNet40 | instance accuracy: 58.51 -> 71.04 | Improved |
| use MOON model-contrastive term instead of FedAvg | PointNet++ SSG · ModelNet40 | instance accuracy: 58.51 -> 45.84 | Hurt |
| use FedNova instead of FedProx | PointNet++ SSG · Craniosynostosis | instance accuracy: 75.83 -> 50.83 | Hurt |
| distill teacher into compact student (74.51% smaller) | SmallPointNet2 · ModelNet40 | instance accuracy: 92.44 -> 92.74 | Improved |
| use Logit-MSE instead of Vanilla KD | SmallPointNet2 · ModelNet40 | instance accuracy: 92.61 -> 87.07 | Hurt |
| keep hard-label cross-entropy term in KD (Logit-MSE) with collapsed SCAFFOLD teacher (ablation) | SmallPointNet2 · ModelNet40 combined | instance accuracy: 8.50 -> 92.94 | Improved |
| remove hard-label cross-entropy (feature/attention objectives) with collapsed teacher (ablation) | SmallPointNet2 · ModelNet40 combined | instance accuracy: 92.94 -> 16.99 | Hurt |
| use chance-level FedAvgM teacher with hard-label Vanilla KD (ablation) | SmallPointNet2 · Craniosynostosis combined | instance accuracy: 99.6 -> 97.5 | No clear effect |
| use chance-level teacher with pure-transfer Attention Transfer (ablation) | SmallPointNet2 · Craniosynostosis combined | instance accuracy: 83.3 -> 25.0 | Hurt |
Closest papers by meaning
- Point-Cache: Test-time Dynamic and Hierarchical Cache for Robust and Generalizable Point Cloud Analysis (CVPR 2025 · similarity 0.85)
- FedVCK: Non-IID Robust and Communication-Efficient Federated Learning via Valuable Condensed Knowledge for Medical Image Analysis (AAAI 2025 · similarity 0.84)
- Federated Learning via Input-Output Collaborative Distillation (AAAI 2024 · similarity 0.84)
- dFLMoE: Decentralized Federated Learning via Mixture of Experts for Medical Data Analysis (CVPR 2025 · similarity 0.84)
- FLUX: Efficient Descriptor-Driven Clustered Federated Learning under Arbitrary Distribution Shifts (NeurIPS 2025 · similarity 0.84)
- Towards Diverse Device Heterogeneous Federated Learning via Task Arithmetic Knowledge Integration (NeurIPS 2024 · 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.
