Research map / Federated learning and privacy / Communication-efficient FL
BAFFLE: A Baseline of Backpropagation-Free Federated Learning
Haozhe Feng, Tianyu Pang, Chao Du, Wei Chen, Shuicheng Yan, Min Lin
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 |
|---|---|---|---|
| increase number of perturbations K (ablation) | BAFFLE · CIFAR-10 (iid, WRN-10-2) | accuracy (%): 52.27 -> 58.63 | Improved |
| increase K from 2000 to 5000 (ablation) | BAFFLE · CIFAR-10 (WRN-10-2) | accuracy: barely improves by 2% | No clear effect |
| add EMA on global parameters (ablation) | BAFFLE · MNIST (WRN-10-2, K=100) | accuracy (%): 85.89 -> 88.35 | Improved |
| use Hardswish instead of ReLU (ablation) | BAFFLE · MNIST (WRN-10-2, K=100) | accuracy (%): 79.08 -> 88.35 | Improved |
| use Hardswish instead of SELU (ablation) | BAFFLE · CIFAR-10 (WRN-10-2, K=200) | accuracy (%): 43.37 -> 52.27 | Improved |
| use twice forward difference instead of central scheme at equal forward cost (ablation) | BAFFLE · CIFAR-10 (WRN-10-2, K=200) | accuracy (%): 42.89 -> 52.27 | Improved |
| label non-iid clients (Dirichlet 0.3, 100 clients) vs iid | BAFFLE · CIFAR-10 / CIFAR-100 | accuracy: drop of 14% on CIFAR-10 and 16% on CIFAR-100 | Hurt |
| increase K on OfficeHome finetuning (ablation) | BAFFLE · OfficeHome (MobileNet) | average Top-1 accuracy (%): 56.91 -> 66.98 | Improved |
| use BAFFLE layer-by-layer inference instead of backpropagation | FL clients · GPU memory CIFAR-10/100 (LeNet) | GPU memory (MB): 1680 -> 67~174 | Improved |
| replace backpropagation with BAFFLE (backprop-free) | FL training · CIFAR-100 (iid, WRN-10-2) | accuracy (%): 60.08 -> 53.15 | Hurt |
Closest papers by meaning
- Low Precision Local Training is Enough for Federated Learning (NeurIPS 2024 · similarity 0.87)
- Efficient Federated Learning against Heterogeneous and Non-stationary Client Unavailability (NeurIPS 2024 · similarity 0.87)
- FedQS: Optimizing Gradient and Model Aggregation for Semi-Asynchronous Federated Learning (NeurIPS 2025 · similarity 0.87)
- FSL-SAGE: Accelerating Federated Split Learning via Smashed Activation Gradient Estimation (ICML 2025 · similarity 0.87)
- Data-Free Black-Box Federated Learning via Zeroth-Order Gradient Estimation (AAAI 2025 · similarity 0.87)
- Adaptive Federated Learning with Auto-Tuned Clients (ICLR 2024 · similarity 0.86)
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.
