Research map / Federated learning and privacy / Communication-efficient FL
Single-Loop Byzantine-Resilient Federated Bilevel Optimization
Yangnan Li, Shenghui Song, Xuanyu Cao
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 single-loop BR-FedBi/BiP/BiM instead of sub-loop BILANTINE | Byzantine-resilient federated bilevel optimization · MNIST hyper-representation (heterogeneous) | worst-case accuracy: Median: 46.02 -> 81.28 (BR-FedBi); RFA: 19.96 -> 70.83 | Improved |
| BILANTINE on CIFAR-10 | BILANTINE baseline · CIFAR-10 hyper-representation | convergence | Hurt |
| add Polyak momentum to BR-FedBi (BR-FedBiM) (ablation) | BR-FedBi · CIFAR-10 hyper-representation (homogeneous) | worst-case accuracy: Median: 43.70 -> 62.67; TM: 42.45 -> 63.18 | Improved |
| add Polyak momentum to BR-FedBi (BR-FedBiM) (ablation) | BR-FedBi · MNIST hyper-representation (heterogeneous), CWMed | worst-case accuracy: 58.25 -> 69.37 | Improved |
| add Polyak momentum (BR-FedBiM) under ALIE with Median aggregator (ablation) | BR-FedBi · MNIST hyper-representation (heterogeneous) | worst-case accuracy: 81.28 -> 79.53 (Median worst case) | Mixed |
| add PAGE variance reduction (BR-FedBiP) to BR-FedBi (ablation) | BR-FedBi · CIFAR-10 hyper-representation (homogeneous) | worst-case accuracy: Median 43.70 -> 59.36; CWTM 42.01 -> 37.70 | Mixed |
Closest papers by meaning
- Distributed Bilevel Optimization with Communication Compression (ICML 2024 · similarity 0.87)
- First-Order Federated Bilevel Learning (AAAI 2025 · similarity 0.87)
- SPARKLE: A Unified Single-Loop Primal-Dual Framework for Decentralized Bilevel Optimization (NeurIPS 2024 · similarity 0.86)
- Efficient Federated Learning against Byzantine Attacks and Data Heterogeneity via Aggregating Normalized Gradients (NeurIPS 2025 · similarity 0.85)
- From Inexact Gradients to Byzantine Robustness: Acceleration and Optimization under Similarity (NeurIPS 2026 · similarity 0.85)
- Locally Differentially Private Decentralized Stochastic Bilevel Optimization with Guaranteed Convergence Accuracy (ICML 2024 · similarity 0.85)
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.
