FedPall: Prototype-based Adversarial and Collaborative Learning for Federated Learning with Feature Drift

Yong Zhang, Feng Liang, Guanghu Yuan, Min Yang, Chengming Li, Xiping Hu, Artificial Intelligence Research Institute, Shenzhen MSU-BIT University, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, School of Medical Technology, Beijing Institute of Technology {zhangyong2023@bit. · 2025

Read FedPall: Prototype-based Adversarial and Collaborative Learning for Federated Learning with Feature Drift by Yong Zhang, Feng Liang, Guanghu Yuan, Min Yang, Chengming Li, Xiping Hu, Artificial Intelligence Research Institute, Shenzhen MSU-BIT University, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, School of Medical Technology, Beijing Institute of Technology {zhangyong2023@bit. (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2507-04781.txt.

Description

Yong Zhang and co-authors improve federated learning, a way for many sites to train one shared AI model without pooling their private data. Each site trains on its own data, which often looks different from the others. When the basic features differ across sites, a problem called feature drift, the shared model learns poorly.

Their framework, FedPall, attacks drift in two ways. A contest-style training step pushes each site's features into a common space, guided by shared class summaries called prototypes. A second step mixes local features with those global summaries to sharpen the final classifier. Tests on three datasets with built-in feature drift show FedPall beat rival methods. The work matters for hospitals and other partners who must learn together without sharing raw data.

Catalog record

CREATORYong Zhang, Feng Liang, Guanghu Yuan, Min Yang, Chengming Li, Xiping Hu, Artificial Intelligence Research Institute, Shenzhen MSU-BIT University, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, School of Medical Technology, Beijing Institute of Technology {zhangyong2023@bit.
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE37 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2507-04781

Browse the full library · About the Tower of Babel