Federated learning (FL) is the emerging technology that adopts collaboratively decentralized privacy-preserving aspects to resolve the impediments of data privacy and security in a distributed framework. The primary aim of federated learning is to perform combined training of machine or deep learning models for optimized real-world applications. The main categorizations of federated learning are horizontal federated learning, vertical federated learning, and transfer federated learning.
Significant characteristics of federated learning are universality for cross-organizational scenarios, enormously non-identically independent distribution (Non-IID), decentralized training, and equality of status. Federated learning is explored in different domains such as healthcare, industrial management, recommender systems, smart cities, autonomous industry, physical information system, natural language processing, edge computing and Internet of Things (IoT), banking, and finance.
Several federated learning surveys investigate the overview, categorization, applications, privacy protection techniques, problems, and future opportunities of federated learning. Statistical heterogeneity, system heterogeneity, data imbalance, resource allocation, and privacy enhancement are considerable challenges of federated learning that need improvement for future scope. Therefore, future development of federated learning requires increased attention and interdisciplinary effort in solving the remained constraints.