通过全局知识蒸馏与局部头部适配的个性化联邦学习
Personalized Federated Learning through Global Knowledge Distillation and Local Head Adaptation
浏览论文内容
中文总结 AI 辅助
针对联邦学习中统计异质性导致单一全局分类器失效的问题，提出pFedKDH方法，仅聚合共享主干并保留客户端特定头部，利用重新校准的全局头部进行知识蒸馏，在多个数据集上取得最优准确率。
中文摘要 AI 辅助
统计异质性限制了联邦学习，当单一全局分类器无法代表客户端特定的标签分布时尤为如此。在本工作中，我们提出了带有头部适配的个性化联邦知识蒸馏（pFedKDH），该方法仅聚合共享主干网络，保留持久的客户端特定头部，并在本地训练期间使用重新校准的全局头部作为教师。在MNIST、Fashion-MNIST、CIFAR10和CIFAR100上，在按类别划分的Dirichlet分区下，pFedKDH在大多数设置中取得了最佳准确率，与最弱基线相比准确率差距高达37.67%，且多次重复实验的标准差持续较低。组件级诊断和收敛结果支持了持久头部和蒸馏引导的本地优化在标签偏斜数据下的作用。
英文摘要
Statistical heterogeneity limits federated learning when a single global classifier cannot represent client-specific label distributions. In this work, we propose Personalized Federated Knowledge Distillation with Head Adaptation (pFedKDH), which aggregates only the shared backbone, keeps persistent client-specific heads, and uses a recalibrated global head as a teacher during local training. Across MNIST, Fashion-MNIST, CIFAR10, and CIFAR100 under class-wise Dirichlet partitions, pFedKDH obtains the best accuracy in most settings, with accuracy gaps up to 37.67\% over the weakest baseline and consistently low standard deviation across repetitions. Component-wise diagnostics and convergence results support the role of persistent heads and distillation-guided local optimization under label-skewed data.
发表机构
- Universidade Federal do Ceará（塞阿拉联邦大学）
- Uppsala University（乌普萨拉大学）
机构由 AI 辅助整理，请以论文原文为准。
