A comprehensive survey of federated transfer learning: Challenges, methods and applications

Federated Learning (FL) has gained significant attention as a novel distributed machine learning paradigm that enables collaborative model training while preserving data privacy. However, traditional FL methods face challenges such as data heterogeneity, system heterogeneity, and labeled data scarcity.

To address these issues, Federated Transfer Learning (FTL), which integrates Transfer Learning (TL) into FL, has attracted the attention of numerous researchers. However, since FL enables a continuous share of knowledge among participants with each communication round while not allowing local data to be accessed by other participants, FTL faces many unique challenges that are not present in TL.

To provide a comprehensive overview of the latest advancements in the field FTL and offer valuable insights for researchers, a research team led by Fuzhen Zhuang published their new research on 15 December 2024 in Frontiers of Computer Science.

In the research, the authors classify the common settings of FTL scenarios into six categories: homogeneous FTL, heterogeneous FTL, dynamic heterogeneous FTL, model adaptive FTL, semi-supervised FTL, and unsupervised FTL. They systematically summarize the solutions to these challenges based on existing FTL works, covering aspects such as motivation, core algorithms, model design, privacy-preserving mechanisms, and communication architectures.

The survey not only covers the theoretical aspects of FTL but also highlights its practical applications. The authors discuss the importance of systems and infrastructure in the success of FTL and outline its current applications across various domains. They also propose future prospects for the development and adoption of FTL.

This comprehensive survey serves as a valuable resource for researchers and practitioners working in the fields of machine learning, data privacy, and distributed computing. It provides a solid foundation for understanding the challenges and opportunities in FTL and is expected to stimulate further advancements in this rapidly evolving field.

Provided by
Higher Education Press