异质稳健分布式支持向量回归

Heterogeneous Robust Distributed Support Vector Regression

摘要: 在分布式算法中，由于本地数据生成机制的异构性，在开发联邦学习方法时考虑个性化非常重要。在这项工作中，本文提出了一种个性化联邦学习方法来解决鲁棒回归问题。具体来说，通过求解具有稀疏融合惩罚的平滑支持向量回归损失来学习回归权重。此外，还设计了用于鲁棒稀疏回归的个性化联邦学习(PerFL-SVR)算法，以有效地解决联邦系统中的估计问题。

Abstract: In distributed algorithms, due to the heterogeneity of local data generation mechanisms, it is crucial to consider personalization when developing federated learning methods. In this work, we propose a personalized federated learning approach to address the robust regression problem. Specifically, the regression weights are learned by optimizing a smoothed support vector regression loss function coupled with a sparse fusion penalty. Furthermore, a personalized federated learning algorithm for robust sparse regression, termed PerFL-SVR, is designed to effectively solve the estimation problem within federated systems.