<tt>PPFL</tt>: A Personalized Federated Learning Framework for Heterogeneous Population

Personalization aims to characterize individual preferences and is widely applied across many fields. However, conventional personalized methods operate in a centralized manner, potentially exposing raw data when pooling individual information. In this paper, with privacy considerations, we develop a flexible and interpretable personalized framework within the paradigm of federated learning, called population personalized federated learning (PPFL). By leveraging “canonical models” to capture fundamental characteristics of a heterogeneous population and employing “membership vectors” to reveal clients’ preferences, PPFL models heterogeneity as clients’ varying preferences for these characteristics. This approach provides substantial insights into client characteristics, which are lacking in existing personalized federated learning (PFL) methods. Furthermore, we explore the relationship between PPFL and three main branches of PFL methods: clustered FL, multitask PFL, and decoupling PFL, and we demonstrate the advantages of PPFL. To solve PPFL (a nonconvex optimization problem with linear constraints), we propose a novel random block coordinate descent algorithm and establish its convergence properties. We conduct experiments in both pathological and practical data sets, and the results validate the effectiveness of PPFL. History: Accepted by Ram Ramesh, Area Editor for Data Science and Machine Learning. Funding: This work was supported by the National Natural Science Foundation for Outstanding Young Scholars of China [Grant 72122018], the National Natural Science Foundation of China [Grant 724B2027], the Humanities and Social Science Fund of the Ministry of Education of China [Grant 22JJD110001], the Shaanxi Provincial Science and Technology Department [Grant 2021JC-01], and the National Key Research and Development Project of China [Grant 2022YFA1004002]. Supplemental Material: The software that supports the findings of this study is available within the paper and its Supplemental Information ( https://pubsonline.informs.org/doi/suppl/10.1287/ijoc.2023.0376 ) as well as from the IJOC GitHub software repository ( https://github.com/INFORMSJoC/2023.0376 ). The complete IJOC Software and Data Repository is available at https://informsjoc.github.io/ .