Cell Handover Model Configuration Using Partial Model Updates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing communication systems waste radio air interface resources when a terminal moves or performs cell handover, as the target access network device redundantly transmits the entire model configuration, leading to inefficiencies and increased overheads.
Innovation Solution
A model configuration method where only the remaining information of the model is sent to the terminal upon cell change, using first indication information to efficiently update the model configuration, reducing redundant transmissions and conserving air interface resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the number of hyperparameters is increased to improve model performance, then model accuracy is improved, but training time and computational resources are consumed
Solution Approach 1:
The patent changes the parameter representation from individual hyperparameter values to distribution parameters (mean and variance) that describe the entire hyperparameter distribution. This transformation reduces the computational complexity of handling numerous hyperparameters while maintaining the ability to capture performance variations through the distribution characteristics.
Solution Approach 2:
The patent creates a simplified copy of the hyperparameter space by representing it through distribution parameters rather than individual hyperparameter instances. This copy captures the essential performance characteristics while enabling faster evaluation and comparison during the search process.
2Measurement precision
If the number of hyperparameters is increased to improve model performance, then model accuracy is improved, but device complexity increases
Solution Approach 1:
The patent transforms the complex multi-dimensional hyperparameter space into a simplified representation using distribution parameters. This parameter transformation reduces system complexity by consolidating multiple hyperparameter dimensions into mean and variance values that can be efficiently managed and compared.
Solution Approach 2:
The patent extracts the essential characteristics of hyperparameter distributions (mean and variance) from the complex hyperparameter space, separating the critical performance-determining factors from the less important details. This extraction simplifies the system while retaining the necessary information for effective hyperparameter optimization.
3Measurement precision
If hyperparameter search space is expanded to find better models, then model performance is improved, but search time increases
Solution Approach 1:
The patent changes the search space from individual hyperparameter values to distribution parameters, enabling efficient exploration of a larger effective search space. By searching over mean and variance parameters, the system can evaluate broader ranges of hyperparameter configurations without proportionally increasing search time.
Data Source
Figure 1a~1b
Figure 2~3a
Figure 3b~3c
AI summary
Embodiments of this application disclose a model configuration method and an apparatus. The method includes: sending, to a terminal that is in a first cell, first-part information of a first model used by the terminal to process data, and sending first indication information that indicates second-part information of the first model, where the second-part information is information to be sent to the terminal when the terminal is located in a second cell, and the first cell and the second cell are different cells. Solutions of embodiments of this application may be widely applied to the field of communication technologies, artificial intelligence, the internet of vehicles, the internet of smart home, or the like.