Federated Learning Group Membership for Dynamic 5G NF Coordination
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The challenge in 5G mobile communication systems is establishing effective federated learning between network functions such as NWDAF and other NFs, or between NWDAF and UE, to enhance network and service intelligence.
Innovation Solution
A method and device for federated learning group processing that involves obtaining characteristic information of a federated learning group, determining a second functional entity based on this information, and dynamically managing the group to improve intelligent performance by accurately adding or removing members.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If federated learning is established between network functions and UEs in 5G systems, then network and service intelligence is enhanced, but the complexity of managing federated learning groups increases
Solution Approach 1:
The patent segments the federated learning management into distinct functional modules: group management network functions that handle group formation and membership, model management functions that handle model training and distribution, and UE-specific management. This segmentation reduces overall system complexity by distributing management responsibilities across specialized components.
Solution Approach 2:
The patent introduces group management network functions as intermediary entities between the core network and UEs. These intermediaries handle the complex tasks of group formation, membership management, and coordination, thereby simplifying the interface between network functions and UEs while enhancing overall system intelligence.
2Adaptability or versatility
If dynamic processing of federated learning groups is implemented, then intelligent performance is improved, but the processing time and resource consumption increase
Solution Approach 1:
The patent implements preliminary actions by pre-establishing group templates, pre-configuring membership criteria, and pre-training baseline models. When dynamic processing is needed, the system can quickly adapt by building upon these pre-prepared elements rather than starting from scratch, thereby reducing processing time while maintaining high intelligent performance.
Solution Approach 2:
The patent implements dynamic group management where federated learning groups can be formed, modified, and dissolved on-demand based on service requirements. The system dynamically adjusts group membership, model training parameters, and resource allocation, enabling flexible adaptation to changing intelligence needs while optimizing processing efficiency through selective dynamic processing.
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
A federated learning group processing method, a device and a functional entity are provided, where the federated learning group processing method includes: obtaining the characteristic information of the federated learning (FL) group; determining the second functional entity according to the characteristic information of the FL group; adding the second functional entity to the FL group.