Research map / Federated learning and privacy / Communication-efficient FL
Server-Proximal Aggregation for Federated Domain-Incremental Learning under Partial Participation: Task-Uniform Convergence and Backward Transfer
Longtao Xu, Jian Li
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 |
|---|---|---|---|
| move proximal anchor from client side to server side (ablation) | SPECIAL · VLCS | ACC (%): 51.17 -> 54.92 | Improved |
| move proximal anchor from client side to server side (ablation) | SPECIAL · PACS | ACC (%): 43.30 -> 45.29 | Improved |
| move proximal anchor from client side to server side (ablation) | SPECIAL · DN4IL | ACC (%): 19.88 -> 24.30 | Improved |
| move proximal anchor from client side to server side (ablation) | SPECIAL · Digit-10 | ACC (%) and BWT (%): ACC 50.61 -> 62.12; BWT -16.42 -> -21.78 | Mixed |
| move proximal anchor from client side to server side (ablation) | SPECIAL · GLUE sequence with DistilBERT | ACC (%): 61.37 -> 74.70 | Improved |
| increase server proximal weight lambda (ablation) | SPECIAL · Digit-10, VLCS, PACS, DN4IL | BWT | Improved |
| increase server proximal weight lambda to very large values (ablation) | SPECIAL · Digit-10, VLCS, PACS, DN4IL | ACC | Hurt |
| set lambda to 0.2 instead of selected 0.25 (ablation) | SPECIAL · Digit-10 | ACC: -0.06% | No clear effect |
| set lambda to 0.3 instead of selected 0.25 (ablation) | SPECIAL · Digit-10 | ACC: -0.6% | Hurt |
| add server-side proximal anchor to FedAvg (SPECIAL vs FedAvg) | FedAvg · DN4IL | ACC (%): 16.47 -> 24.30 | Improved |
| add server-side proximal anchor to FedAvg (SPECIAL vs FedAvg) | FedAvg · Digit-10 | BWT (%): -40.44 -> -21.78 | Improved |
| use SPECIAL instead of replay-based methods | SPECIAL · Digit-10, VLCS, PACS, DN4IL | ACC (%): 62.12 vs 61.63 (SR-FDIL) on Digit-10 | Improved |
Closest papers by meaning
- HERO: A Heterogeneity-Aware Benchmark Library for Federated Continual Learning (NeurIPS 2026 · similarity 0.89)
- Exact and Linear Convergence for Federated Learning under Arbitrary Client Participation is Attainable (NeurIPS 2025 · similarity 0.89)
- Towards Efficient Replay in Federated Incremental Learning (CVPR 2024 · similarity 0.88)
- FedSOL: Stabilized Orthogonal Learning with Proximal Restrictions in Federated Learning (CVPR 2024 · similarity 0.88)
- Federated Learning under Periodic Client Participation and Heterogeneous Data: A New Communication-Efficient Algorithm and Analysis (NeurIPS 2024 · similarity 0.88)
- Efficient Federated Learning against Heterogeneous and Non-stationary Client Unavailability (NeurIPS 2024 · similarity 0.88)
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.
