Federated Learning Group Selection for Dynamic 5G Network Functions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The challenge of establishing federated learning between network functions (such as NWDAF and other NFs) or between network functions and UEs in the 5G mobile communication system has not been effectively addressed.
Innovation Solution
A method 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 adding it to the group, including processes for joining and leaving the group based on specific conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If federated learning groups are established without dynamic management mechanisms, then initial group formation is simple, but the system cannot adapt to changing network conditions and service requirements
Solution Approach 1:
The patent implements dynamic management of federated learning groups by establishing mechanisms for real-time addition and removal of network functions based on changing network conditions, service requirements, and performance metrics. The system continuously evaluates whether NFs should be added or removed from FL groups, enabling adaptation to dynamic environments while maintaining structured management through defined procedures and criteria.
2Productivity
If all network functions are included in federated learning groups, then comprehensive data analysis is achieved, but processing efficiency decreases and resource consumption increases
Solution Approach 1:
The patent segments the network functions into multiple federated learning groups based on service types, network domains, and functional characteristics. This segmentation allows each FL group to process data independently and efficiently, reducing the computational burden on individual groups while maintaining comprehensive analysis capabilities across the entire network through coordinated operation of multiple specialized groups.
Solution Approach 2:
The patent implements selective participation mechanisms where only network functions that meet specific criteria and have relevant data are included in particular FL groups. This partial participation approach ensures that each FL group contains only the necessary subset of NFs required for specific analysis tasks, improving processing efficiency while avoiding the overhead of including all possible network functions in every group.
3Reliability
If strict criteria are enforced for joining federated learning groups, then group quality and relevance are maintained, but the flexibility to incorporate new network functions is reduced
Solution Approach 1:
The patent establishes pre-defined evaluation criteria and qualification standards for network functions before they are added to federated learning groups. These preliminary criteria cover data quality, computational capabilities, security requirements, and service relevance. By setting these standards in advance, the system ensures that only qualified NFs are considered for inclusion, maintaining group quality while providing a clear, systematic pathway for new functions to be evaluated and incorporated when they meet the established requirements.
Data Source
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.