Model Accuracy Provisioning Across NWDAF Analytics Transfer
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current standard specifications fail to address the handling of model accuracy information for machine learning models during analytics transfers between network entities, leading to misconfigurations, unnecessary signaling, and inaccurate retraining or model selection due to differing configurations between source and target network functions.
Innovation Solution
A network entity generates and provides indications for changing or relocating model accuracy information provisioning, ensuring seamless consumption by another entity, using indications such as termination, relocation, or changes in provisioning, and registrations, compatible with existing standards.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If analytics ID is transferred from source NWDAF to target NWDAF, then service continuity is improved, but model accuracy information consumption is interrupted
Solution Approach 1:
The source NWDAF proactively notifies the consuming NF about the analytics transfer before or during the transfer process. This preliminary action ensures that the consumer can update its configuration to continue receiving model accuracy information from the target NWDAF, preventing any interruption in information consumption while maintaining service continuity
Solution Approach 2:
The system establishes a feedback mechanism where the source NWDAF informs the consumer NF about the transfer of analytics ID to target NWDAF. This feedback loop enables the consumer to adjust its information consumption path, ensuring continuous access to model accuracy information despite the analytics transfer between network entities
2Adaptability or versatility
If target NWDAF uses different configuration than source NWDAF, then local optimization is improved, but model accuracy information accuracy deteriorates
Solution Approach 1:
The source NWDAF provides information about the transferred analytics ID and associated ML model to the consumer NF before the transfer is complete. This preliminary information transfer ensures that the consumer can properly configure itself to receive accurate model accuracy information from the target NWDAF, maintaining measurement precision despite configuration differences between source and target NWDAFs
Solution Approach 2:
The consumer NF acts as an intermediary that receives notification from the source NWDAF about the analytics transfer. This intermediary role allows the consumer to bridge the configuration gap between source and target NWDAFs, ensuring continuous and accurate reception of model accuracy information regardless of the target NWDAF's different local configuration
3Loss of energy
If ML model accuracy information is not transferred to target NWDAF, then signaling overhead is reduced, but model retraining accuracy deteriorates
Solution Approach 1:
The patent extracts only the essential information needed for model accuracy consumption from the source NWDAF to the consumer NF, rather than transferring all analytics data. By selectively extracting and transferring only the necessary model accuracy information and transfer notifications, the system minimizes signaling overhead while ensuring the consumer has sufficient information to maintain accurate model retraining and selection
Solution Approach 2:
The consumer NF is empowered to autonomously manage its information consumption by receiving notifications about analytics transfers. The consumer can independently update its configuration to continue consuming model accuracy information from the target NWDAF without requiring explicit retraining data transfer, reducing signaling overhead while maintaining retraining accuracy through self-service information management
Data Source
AI summary
Model accuracy information for a model, which is associated with a model identifier (ID) and/or an analytics ID, is provided and consumed in a mobile communication network having a plurality of network entities. The model may be used for generating and/or providing analytics information for the analytics ID. Continuous consumption of the model accuracy information is enabled, even in the case of an analytics transfer, including, for instance, a transfer of the analytics ID from one network entity to another. The network entities are configured to respectively provide and receive indications to change a provisioning of model accuracy information and indications related to a change of the provisioning of model accuracy information.