AI Model Layer Synchronization to Reduce Transmission Resources
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The synchronization of updated AI network models between transmit and receive ends in a communication network consumes substantial transmission resources, reducing resources available for other information transmission.
Innovation Solution
Implementing transfer learning to update only specific network layers of the AI network model, synchronizing only the parameter information of these updated layers instead of the entire model, thereby reducing resource consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the entire AI network model is synchronized between transmit and receive ends, then the model update completeness is improved, but the transmission resource consumption increases
Solution Approach 1:
The patent segments the AI network model into multiple network layers and identifies only the updated target network layer for synchronization. Instead of transmitting the entire model, the system divides the model structure and selectively updates specific segments (layers), thereby reducing transmission resource consumption while maintaining model effectiveness.
Solution Approach 2:
The patent extracts only the necessary parameter information from the target network layer that has been updated through transfer learning. By taking out only the essential update parameters rather than the complete model, the system reduces transmission overhead while ensuring the receive end can effectively update its local model.
2Adaptability or versatility
If air interface-based transmission is used for model synchronization, then the model can be updated at both ends, but transmission resources for other information are reduced
Solution Approach 1:
The patent applies partial action by synchronizing only the parameter information of the target network layer rather than the entire AI model. This partial update approach maintains the adaptability of model synchronization while significantly reducing the quantity of transmission resources consumed, leaving more resources available for other information transmission.
3Loss of energy
If transfer learning is implemented to update only specific network layers, then transmission resource consumption is reduced, but the complexity of model update management increases
Solution Approach 1:
The patent performs preliminary action by pre-processing the model update to identify which network layer is the target layer before transmission. The network-side device determines the target network layer in advance and prepares only the necessary parameter information for transmission, which simplifies the overall update management process despite the selective update approach.
Solution Approach 2:
The patent implements feedback mechanisms where the network-side device sends indication information to the terminal about the target network layer, and the terminal uses this feedback to accurately update its local model. This feedback loop ensures that both ends maintain consistency in model updates while managing complexity through structured communication.
Data Source
AI summary
This application discloses an information transmission method, an information transmission apparatus, and a communication device. The information transmission method of embodiments of this application includes: transmitting, by a terminal, first information, where the first information includes at least one of the following: parameter information of a target network layer, where the target network layer is an updated network layer in a target AI network model; and second information, where the second information is used to describe the target network layer.