ML Model Broker for 5G Network Discovery
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In 5G systems, consumer Network Functions (NFs) face difficulties in identifying and accessing the most appropriate machine learning models and NWDAF instances for specific analytics scenarios due to cumbersome discovery mechanisms.
Innovation Solution
A broker component maintains a provider register with information about machine learning model providers and their models, enabling it to match consumer requests with available models based on characteristics such as output parameters, input parameters, model types, and evaluation metrics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If consumer NFs use the traditional discovery mechanisms provided by the Network Repository Function to find NWDAF instances and models, then the system can maintain flexibility in model selection, but the process becomes cumbersome and complex making it difficult to identify the most appropriate model
Solution Approach 1:
The patent introduces a broker component as an intermediary between consumer NFs and NWDAF instances. This broker maintains a registry of available models and their characteristics, and facilitates the matching process by receiving model requests from consumers, querying the registry for suitable models, and returning matching results. This intermediary approach simplifies the discovery process for consumers while maintaining system flexibility.
2Adaptability or versatility
If multiple NWDAF instances provide different models for the same analytics, then model diversity and specialization are improved, but it becomes difficult for consumer NFs to identify and select the most appropriate model
Solution Approach 1:
The patent employs characteristic attributes as identifying markers for different models, similar to color changes. Each model is described with specific characteristics such as output parameters, input parameters, model types, and evaluation metrics. These characteristics serve as distinctive identifiers that allow consumer NFs to easily distinguish between different models and select the most appropriate one based on their specific needs.
Solution Approach 2:
The broker component provides feedback to consumer NFs by returning matching model information based on their requests. The system queries the registry with the desired model characteristics and provides feedback with suitable model options, enabling consumers to identify and select appropriate models efficiently.
3Measurement precision
If consumer NFs manually query multiple NWDAF instances to find suitable models, then model selection accuracy can be improved, but the time and resources required for discovery increase
Solution Approach 1:
The broker component performs preliminary action by pre-querying the registry with the desired model characteristics before the consumer NF needs to select a model. The broker retrieves and filters suitable models in advance, so when the consumer NF makes a request, the matching results are already prepared and available for immediate return, significantly reducing discovery time while maintaining selection accuracy.
Data Source
Figure 1~2
Figure 3
Figure 4a~4c
AI summary
A technique for facilitating use of machine learning models in a system comprising a plurality of machine learning model providers (802) is disclosed. A method implementation of the technique is performed by a broker component (804) maintaining a provider register containing information about the plurality of machine learning model providers (802) and machine learning models provided by the plurality of machine learning model providers (802). The method comprises receiving a request for a desired machine learning model from a machine learning model consumer (902), determining, based on the information contained in the provider register, a machine learning model among the machine learning models provided by the machine learning model providers (802) that matches the desired machine learning model, and sending a response to the machine learning model consumer (902) providing information associated with the determined machine learning model.