用生成模型解决极端非独立同分布下的联邦学习难题
FederatedFactory: Generative One-Shot Learning for Extremely Non-IID Distributed Scenarios
- 将联邦学习单位从参数改为生成模块，实现一次通信合成均衡数据
- 在极端异构场景下，准确率从11.36%提升至90.57%，恢复医疗图像识别性能
- 无需预训练模型，支持可逆删除生成模块，适合隐私敏感场景
联邦学习（FL）可在不泄露数据的前提下实现分布式优化。然而，在本地标签分布互斥的情况下，标准权重聚合因优化方向冲突而失效。现有方法常依赖预训练基础模型，假设不切实际。我们提出FederatedFactory，一个零依赖框架，将联邦单元从判别参数转为生成先验。通过单轮通信交换生成模块，该架构可原生合成全局类平衡数据集，彻底消除梯度冲突与外部先验偏差。在多个医学影像基准测试中，包括MedMNIST和ISIC2019，我们的方法恢复了集中式最优性能。在病理级异构条件下，基线准确率从11.36%提升至CIFAR-10的90.57%，并将ISIC2019的AUROC恢复至90.57%。此外，该框架可通过确定性删除特定生成模块实现精确模块化遗忘。
原文摘要 · Abstract (English)
Federated Learning (FL) enables distributed optimization without compromising data sovereignty. Yet, where local label distributions are mutually exclusive, standard weight aggregation fails due to conflicting optimization trajectories. Often, FL methods rely on pretrained foundation models, introducing unrealistic assumptions. We introduce FederatedFactory, a zero-dependency framework that inverts the unit of federation from discriminative parameters to generative priors. By exchanging generative modules in a single communication round, our architecture supports ex nihilo synthesis of universally class balanced datasets, eliminating gradient conflict and external prior bias entirely. Evaluations across diverse medical imagery benchmarks, including MedMNIST and ISIC2019, demonstrate that our approach recovers centralized upper-bound performance. Under pathological heterogeneity, it lifts baseline accuracy from a collapsed 11.36% to 90.57% on CIFAR-10 and restores ISIC2019 AUROC to 90.57%. Additionally, this framework facilitates exact modular unlearning through the deterministic deletion of specific generative modules.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。