Federated Learning WTRU Consent for Efficient Participant Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing federated learning (FL) systems face inefficiencies in selecting wireless transmit/receive units (WTRUs) for AI/ML operations, leading to resource waste due to unnecessary verification of large numbers of WTRUs, and lack of consideration for diversity and efficiency in member selection.
Innovation Solution
A network node receives consent indications and service capabilities from WTRUs, sending registration responses and notifications for AI/ML operation enhancements, enabling efficient selection of WTRUs for FL tasks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a 5G network checks or verifies whether a large number of WTRUs may participate in an application layer AI/ML training operation, then the selection of members for AI/ML operations can satisfy requirements for diversity, but resource waste occurs due to the large number of verifications required
Solution Approach 1:
The patent applies preliminary action by having WTRUs proactively indicate their consent to participate in AI/ML operations before being selected. The network node receives and stores these consent indications in advance, so when AI/ML training is needed, the network can immediately query the stored consent information without performing real-time verification of each WTRU. This preliminary collection of consent data eliminates the resource-wasting verification process while maintaining the ability to select diverse participants from the pre-qualified pool.
2Reliability
If the network verifies consent for each WTRU individually during AI/ML operations, then accurate selection is achieved, but the verification process becomes time-consuming and reduces operational efficiency
Solution Approach 1:
The network node performs the consent verification action in advance by receiving and storing consent indications from WTRUs before AI/ML operations commence. When training operations are initiated, the network queries this pre-stored consent information rather than performing individual verification during the operation. This shifts the verification process to a preliminary stage, ensuring accurate selection of consenting WTRUs while eliminating time delays during actual AI/ML operations.
Solution Approach 2:
The patent implements copying by creating and storing a copy of the consent indication data in the network node. Instead of retrieving or verifying original consent data from each WTRU during operations, the network uses these stored copies for rapid querying and selection. This copying mechanism maintains selection accuracy while dramatically reducing the time required for member selection during AI/ML operations.
3Productivity
If WTRUs are selected for FL training without considering their service capabilities, then the selection process is simple, but efficiency of the training operation deteriorates
Solution Approach 1:
The patent applies local quality by having different WTRUs indicate their specific service capabilities (such as computational power, available resources, or specialized functions) in their consent indications. The network node stores these capability-specific details and uses them to make targeted selections for AI/ML training operations. This ensures that WTRUs with appropriate local qualities (capabilities) are selected for specific training tasks, improving overall training efficiency without requiring a complex centralized evaluation process for each operation.
Data Source
AI summary
A method and apparatus for supporting artificial intelligence/machine learning (AI/ML) operating are provided herein. A network node may receive a request to subscribe for notifications indicating consent provided by one or more wireless transmit receive units (WTRUs) to AIML operation supporting service enhancement. The network node may receive a registration request message including an indication that the WTRU provides consent to AI/ML operation supporting at least one type of service enhancement, and a service capability container indicating one or more service capabilities of the WTRU. The network node may send a registration response message indicating that a consent status associated with the WTRU is active for one or more applications. The network node may send a notification message indicating consent provided by the randomly selected WTRU to AI/ML operation supporting service enhancement for at least one of the one or more applications.