Terminal AI Model Transfer Categories for Low-Overhead Inference
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing radio communication technologies lack an efficient method for transferring AI models and data sets between terminals and base stations, leading to increased communication overhead and inhibited communication throughput and quality due to the lack of defined transfer methods and interoperability between vendors.
Innovation Solution
A terminal and base station system that includes a receiving section for identifying AI model transfer categories and a control section for performing inference based on the identified models, utilizing model transfer categories to minimize overhead and enhance communication efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If AI models and data sets are transferred between terminals and base stations, then advanced AI functions can be implemented, but communication overhead increases
Solution Approach 1:
The patent segments AI model transfer into different categories (full model transfer, partial model transfer, and data set transfer) based on specific needs. This allows selective transfer of only necessary model components or data, reducing unnecessary communication overhead while maintaining the capability to implement advanced AI functions.
Solution Approach 2:
The patent introduces model transfer categories as a parameter to control the transfer process. By changing the category parameter, the system can adaptively select the appropriate transfer mode (full model, partial model, or data set), thereby optimizing communication overhead based on specific operational requirements.
2Adaptability or versatility
If AI model transfer methods are not appropriately defined, then vendor interoperability cannot be achieved, but communication throughput and quality are inhibited
Solution Approach 1:
The patent establishes a universal model transfer framework with standardized categories that can be applied across different vendors and scenarios. This universal approach enables vendor interoperability while maintaining efficient communication throughput through standardized procedures for full model transfer, partial model transfer, and data set transfer.
3Reliability
If complete AI models are transferred, then model accuracy is maintained, but communication overhead and transfer time increase
Solution Approach 1:
The patent segments the AI model into different transfer categories: full model transfer for scenarios requiring complete accuracy, and partial model transfer for scenarios where only specific components are needed. This segmentation allows the system to maintain model accuracy when necessary while reducing transfer time by transferring only essential components when full accuracy is not required.
Solution Approach 2:
The patent applies partial action by introducing partial model transfer, where only the necessary portions of the AI model are transferred instead of the complete model. This reduces transfer time and communication overhead while providing sufficient accuracy for specific applications, avoiding the excessive transfer of unnecessary model components.
Data Source
AI summary
A terminal according to one aspect of the present disclosure includes a receiving section that receives information for identification of an artificial intelligence (AI) model based on a model transfer category, and a control section that performs inference based on the AI model identified based on the information. According to one aspect of the present disclosure, appropriate inference/use of resources based on preferable overhead reduction/AI model can be implemented.