Network-Orchestrated AI Model Training With QoS-Based Execution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing communication systems lack the capability to provide endogenous intelligence services such as AI model training.
Innovation Solution
A model training method and apparatus that enables communication systems to provide AI model training services by interacting among multiple entities to determine QoS parameters and execution operations, allowing for closed-loop model training on demand.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If existing communication systems are used, then connection and data transmission services can be provided, but endogenous intelligence services such as AI model training cannot be provided
Solution Approach 1:
The communication system is extended to perform multiple functions by integrating AI model training capabilities alongside existing connection and data transmission services. The network device acts as a universal platform that can both transmit data and execute model training operations, allowing a single system to serve diverse purposes including traditional communication and emerging AI services
2Adaptability or versatility
If AI model training service is added to communication system, then endogenous intelligence capability is improved, but system complexity increases
Solution Approach 1:
The AI model training process is segmented into distinct operational phases including parameter determination, message sending, and result receiving. The system divides the training task into manageable components that can be executed sequentially, with the network device handling specific segments like QoS parameter determination and execution operation management, thereby reducing overall system complexity through structured decomposition
Solution Approach 2:
A first entity (network device) is introduced as an intermediary between the second entity (service requester) and the third entity (model training executor). This intermediary manages the complexity by handling QoS parameter determination, coordinating execution operations, and facilitating communication between other components, thereby shielding the system from complexity while enabling AI model training functionality
3Adaptability or versatility
If multiple entities interact for model training, then service flexibility is improved, but communication overhead increases
Solution Approach 1:
QoS parameters and execution operations are determined in advance before the actual model training execution. The first entity pre-processes training requirements, determines optimal parameters, and prepares execution plans beforehand. This preliminary action reduces communication overhead during actual training by having critical decisions made beforehand, allowing the system to maintain flexibility while minimizing real-time communication needs
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
The present disclosure relates to the technical field of communications. Provided are a model training method and apparatus, and a storage medium. The method comprises: a second entity sending a model training request message to a first entity; the first entity determining model-training-related information according to the model training request message, and sending a first indication message to a third entity; the third entity sending an execution operation request message to at least one sixth entity according to a target QoS index parameter list and an execution operation list; each sixth entity performing model training according to the execution operation request message, so as to obtain a model training result, and sending an execution operation response message to the third entity, wherein the execution operation response message comprises the model training result; the third entity sending a first response message to the first entity; and the first entity sending a second response message to the second entity, wherein the first response message and the second response message each comprise the model training result. By means of interaction among a plurality of entities, a communication system can provide a model training service for a user, a network or a third party as required.