Federated Learning: A Comprehensive Survey of Applications, Challenges, and Emerging Research Frontiers
جاري التحميل...
التاريخ
عنوان الدورية
ردمد الدورية
عنوان المجلد
الناشر
th IEEE International Conference on Electronic Engineering
خلاصة
Federated Learning (FL) is seen to have revolutionized collaborative machine learning, wherein models are trained on distributed resources without compromising data privacy. This survey gives an impression of the FL landscape and analyzes its various architectures: horizontal, vertical, and federated transfer learning, each targeting different data distribution scenarios. We detail FL's ever-increasing applications in prominent areas: better diagnostics and drug discovery in healthcare, fraud prediction and risk estimation in finance, improved traffic and energy management in smart cities, intelligent edge computing for IoT devices, and so forth. While FL holds great potential, it also faces a tangled web of challenges. We take a close look at the widespread issue of data heterogeneity, where varying data distributions among clients can hinder the global model's ability to converge and perform well. In addition, we address the heavy communication costs involved in repeated model exchanges and explore advanced techniques such as differential privacy and secure multi-party computation to safeguard sensitive information. Lastly, there are scaling issues discussed in FL as it grows to accommodate a large number of clients, highlighting the need for robust aggregation schemes and strong system designs. In this work, we explore where Federated Learning currently stands, unpack its layered challenges, and point toward potential paths that researchers could pursue next.