ORDERS：面向个性化联邦学习的范数秩聚合实证研究
ORDERS: An Empirical Study of Norm-Rank Aggregation for Personalized Federated Learning
Koffka Khan
中文摘要
个性化联邦学习在共享表征与客户端专属预测器之间折中，但服务器加权规则的实际贡献往往被本地训练与评估选择所掩盖。研究 ORDERS 配置，结合共享骨干网络、私有残差适配器与分类器、按更新范数降序分配的几何权重、特征对齐，以及私有参数扰动。服务器仅对同一广播模型的各路更新做加权和，不通过顺序累加获得额外优化效应。完整评估共 80 次最终运行：8 种配置、2 个数据集、每数据集固定切分下的 5 个训练种子。在每客户端两类的 CIFAR-10 上，ORDERS 达到 80.51 ± 0.79% 的原生客户端平均精度，FedPer-R1 为 79.02 ± 1.42%，匹配的均匀权重对照组为 80.27 ± 0.73%；经过通用本地微调后，与 FedPer-R1 的差距缩小至 0.32 个百分点。在 Sent140 上，ORDERS 达到 74.71 ± 0.49%，仅高于事后客户端训练多数投票基线 0.69 个百分点。消融实验仅在端点处提供有限、依赖终态的证据支持范数秩与对齐，扰动未带来明显收益。参数载荷开销分别为 5.47% 与 0.78%。
关键要点
- 01问题：服务器加权规则在个性化联邦学习中的贡献常被本地训练与评估选择掩盖，难以纯化度量。
- 02方法：ORDERS = 共享骨干 + 私有残差适配器与分类器 + 按更新范数降序的几何权重 + 特征对齐 + 私有参数扰动。
- 03结果：CIFAR-10（每客户端两类）原生精度 80.51 ± 0.79%，略优于均匀权重对照；Sent140 达 74.71 ± 0.49%。
- 04局限：消融仅在端点处提供有限证据，范数秩与对齐的收益依赖终态，扰动无明显收益。
- 05开销：相比共享基线，参数载荷分别增加 5.47%（CIFAR-10）与 0.78%（Sent140）。
解读
尚无解读。
原始英文摘要
arXiv:2610.10361v1 Announce Type: new Abstract: Personalized federated learning combines shared representations with client-specific predictors, but the contribution of a server weighting rule can be obscured by local training and evaluation choices. We study ORDERS, a configuration that combines a shared backbone, a private residual adapter and classifier, geometric weights assigned by descending update norm, feature alignment, and private-parameter perturbations. The server computes a weighted sum of updates obtained from the same broadcast model; it does not obtain an additional optimization effect from sequential addition. A fully specified evaluation comprises 80 final runs: eight configurations, two datasets, and five training seeds on one fixed partition per dataset. On two-class-per-client CIFAR-10, ORDERS achieves $80.51 \pm 0.79\%$ native mean client accuracy, compared with $79.02 \pm 1.42\%$ for FedPer-R1 and $80.27 \pm 0.73\%$ for the matched uniform-weight control. After common local fine-tuning, the difference from FedPer-R1 narrows to 0.32 percentage points. On Sent140, ORDERS reaches $74.71 \pm 0.49\%$, only 0.69 points above a post hoc client training-majority diagnostic. Ablations provide limited, endpoint-dependent evidence for norm ranking and alignment, and no clear benefit from perturbations. Parameter-payload savings are 5.47% and 0.78%, respectively.