Communication AI Model Selection by Scenario Granularity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing communication systems face challenges in ensuring AI model performance across different scenarios due to varying computing capabilities of devices, leading to inefficient model switching overheads and suboptimal generalization, as predefined scenario classifications do not account for device-specific capabilities.
Innovation Solution
A method for a first communication device to determine an association relationship between scenario and configuration information to customize the scenario classification granularity, allowing it to train or select AI models based on device-specific conditions, thereby optimizing model usage and reducing overheads.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a fixed scenario classification is used to train AI models for each communication scenario, then the communication device can use different AI models in different scenarios with small model scales, but additional model switching overheads are caused
Solution Approach 1:
The patent applies dynamics by making the scenario classification granularity adjustable rather than fixed. The system dynamically determines the classification granularity based on device computing capabilities and operational needs, allowing the AI model selection strategy to adapt to changing conditions and resolve the contradiction between model performance and switching overheads
Solution Approach 2:
The patent changes the parameter of scenario classification granularity from a fixed value to a variable that can be adjusted according to device capabilities. By modifying this key parameter, the system optimizes the balance between training multiple specialized models and using a single model with different classification granularities, thereby reducing model switching overheads while maintaining performance
2Quantity of substance
If a fixed scenario classification is used, then the AI model scale can be kept small for each scenario, but a device having strong computing capability cannot use an AI model with larger scale and stronger generalization capability
Solution Approach 1:
The system dynamically adjusts the scenario classification granularity based on the specific device's computing capabilities. Devices with stronger computing capabilities can utilize finer or coarser classification granularities to access models with different scales, enabling them to leverage larger models with stronger generalization capabilities when needed, thus resolving the contradiction between model quantity and device adaptability
Solution Approach 2:
The patent creates a universal framework where a single AI model can serve multiple functions by adapting to different scenario classifications with varying granularities. This allows one model to be used across different devices with different capabilities, making the system more versatile and eliminating the need for separate models for each device type
3Ease of manufacture
If a fixed scenario classification is used, then the model training process is simplified, but different devices cannot customize their own scenario classification granularity based on computing capabilities
Solution Approach 1:
The patent introduces dynamic customization of scenario classification granularity based on device capabilities. Instead of a fixed classification scheme, the system allows each device to configure its own appropriate granularity level, enabling devices with different computing capabilities to train and use models that are optimized for their specific constraints while maintaining a relatively simple overall training process
Data Source
AI summary
Embodiments of this application provide a model usage method and a related device, and relate to the communication field. In the method, a first communication device may determine a first association relationship based on a status of the first communication device, to customize a scenario classification granularity. Then, the first communication device may obtain first information from a second communication device to train at least one first AI model in M first AI models, or the first communication device may determine one first AI model for inference from M first AI models based on the first association relationship. The first association relationship is an association relationship between X pieces of scenario information, Y pieces of configuration information, and the M first AI models of the first communication device, X, Y, and M are positive integers, and M is less than or equal to X*Y.