Network ML Model Repository for Distributed 5G Model Updates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing telecommunication networks face challenges in efficiently managing machine learning models due to their geographical spread, leading to difficulties in training, storing, and distributing these models effectively across target network elements.
Innovation Solution
A system comprising a central module and edge modules implemented as network functions in a core network, such as a 5G core, manages machine learning models through storage and distribution, ensuring up-to-date models are utilized by target network elements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If machine learning models are distributed across geographically spread network elements, then network service coverage is improved, but model management complexity increases
Solution Approach 1:
The system segments model management into two distinct components: a centralized model repository function that stores and manages machine learning models, and distributed network elements that consume these models. This segmentation allows the network to maintain wide service coverage through distributed elements while the centralized repository handles the complexity of model training, versioning, and distribution.
Solution Approach 2:
The patent introduces a centralized model repository function as an intermediary between model training sources and distributed network elements. This intermediary manages the complexity of model lifecycle operations (training, validation, versioning) and provides a standardized interface for model distribution, thereby reducing management complexity across the geographically distributed network.
2Reliability
If machine learning models are updated frequently to improve performance, then model accuracy is improved, but distribution time and network load increase
Solution Approach 1:
The centralized model repository function performs preliminary actions by training, validating, and preparing models for distribution before they are needed by network elements. Models are pre-processed and staged in the repository, allowing for efficient distribution when updates are required without causing network delays.
Solution Approach 2:
The system implements feedback mechanisms where network elements report model performance and requirements to the centralized repository. This feedback enables the repository to determine when model updates are necessary, optimizing the balance between maintaining high model accuracy and minimizing unnecessary distribution operations that would consume network resources.
3Stability of the object's composition
If a centralized system manages all machine learning models, then model consistency is improved, but system single point of failure risk increases
Solution Approach 1:
While maintaining a centralized model repository for consistency, the system allows distributed network elements to have local model caches and validation capabilities. This local quality ensures that individual network elements can continue operating with cached models if the centralized repository becomes unavailable, thereby reducing the single point of failure risk while maintaining model consistency during normal operations.
Data Source
AI summary
Provided are system, method, and device for managing machine learning models in a network. According to embodiments, the system may include: a memory storage storing computer-executable instructions; and at least one processor communicatively coupled to the memory storage, wherein the at least one processor may be configured to execute the instructions to: store one or more machine learning model; and distribute the one or more machine learning model to one or more target network element in a network.