Model management method and device, communication equipment, storage medium and program product
By introducing a first network element into the communication system and utilizing user-selectable configuration and address management, the problems of high storage overhead and poor generalization of AI models on the base station side are solved, achieving efficient training and improved generalization performance of AI models.
Patent Information
- Application Number
- CN202510044948.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-10
- Publication Date
- 2026-07-10
AI Technical Summary
In AI physical layer use cases, AI model management on the base station side suffers from high storage overhead and poor generalization, especially when user configuration changes or cell handover requires frequent model switching.
The first network element is introduced, which guides the UE to perform channel measurement by sending the user selection configuration and first address to N base stations. The training dataset is then routed to the first network element for clustering and model training, and a highly generalizable AI model is distributed to the base stations.
It effectively reduces the storage and training dataset acquisition costs of AI models, improves the generalization performance and adaptability of AI models, and reduces the frequency of model switching.