Call for Papers: Federated Learning for Cognitive Network Management
IEEE Communications Magazine · 2022
BackgroundRecent years have witnessed explosive growth in using Supervised, Un-Supervised, Semi-supervised, Reinforcement, and Deep Reinforcement Learning to solve networking issues.Nevertheless, since next-generation networks are complex, dynamic, and non-centralized by nature, it is worth exploring Federated Learning and its variations to cognitive network management set as making automated decisions for management actions through autonomous, zero-touch, self-driven, and knowledge-driven techniques.Federated Learning is a particular distributed machine learning approach.Distributed machine learning algorithms create accurate models using multiple servers, usually containing datasets of around the same size with independent and identically distributed samples, aiming to improve the learning process regarding time, memory, and bandwidth.Federated Learning algorithms hold the potential of becoming one of the leading 6G enablers since they can build accurate models from vast decentralized and heterogeneous datasets on resource-constrained devices (e.g., gateways, edge devices, smartphones, and autonomous vehicles).Furthermore, the Federated Learning process can be coordinated by centralized nodes (e.g., 5G/6G network data analytics functions and SDN controllers) or collaboratively by distributed nodes (e.g., in-slice managers and programmable switches).Federated Deep Reinforcement Learning and Online Federated Learning have been proposed recently for leveraging the potential of combining Federated Learning, Deep Reinforcement Learning, and Online Learning.In these combinations, learning agents would learn deeply and continuously by interacting with the environment to meet eXperience Level Agreements and Service Level Agreements in 5G, 6G, and datacenter networks, for instance.This Feature Topic (FT) aims to bring together researchers, industry practitioners, and individuals working on the related areas to share their new ideas, latest findings, and state-of-the-art results.