Machine Learning Model Update via Pre-trained Mapping in Communication Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In communication systems, machine learning models require frequent retraining due to environmental changes, which is inefficient and not suitable for high-promptness requirements, affecting system performance.
Innovation Solution
A method and apparatus that utilize pre-established mapping relationships between communication environment information and machine learning models to quickly update deployed models on access network devices and terminals, ensuring suitability for current environments without retraining.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If machine learning models are retrained multiple times to adapt to communication environment changes, then model performance is maintained, but update efficiency deteriorates due to the time-consuming retraining process
Solution Approach 1:
The patent pre-trains multiple machine learning models during the off-line stage with different communication environment characteristics before deployment. When the communication environment changes, the system can directly select and switch to a pre-trained model that matches the new environment, eliminating the need for time-consuming on-line retraining and enabling rapid model updates while maintaining performance.
2Adaptability or versatility
If machine learning models are retrained frequently to adapt to environment changes, then model suitability is improved, but system time consumption increases
Solution Approach 1:
Multiple machine learning models are pre-trained in advance with different communication environment characteristics and stored in the model pool. When environment changes occur, the system performs rapid model selection and switching based on current environment matching, avoiding time-consuming retraining processes and enabling fast adaptation to new environments.
Solution Approach 2:
The patent changes the approach from modifying model parameters through retraining to selecting different pre-trained models with different parameter configurations. By maintaining multiple models with varying parameters in the model pool, the system can adapt to environment changes by switching between models rather than retraining, significantly reducing time consumption.
3Reliability
If machine learning models are retrained to adapt to new communication environments, then model effectiveness is improved, but retraining complexity increases
Solution Approach 1:
Multiple machine learning models are pre-trained during the off-line stage with different communication environment characteristics and deployed to the model pool. When the communication environment changes, the system can directly select and switch to an appropriate pre-trained model, avoiding the need for complex on-line retraining processes and reducing system complexity.
Solution Approach 2:
The patent creates multiple copies of machine learning models trained with different characteristics and stores them in the model pool. Instead of retraining a single model when the environment changes, the system copies the functionality by selecting from pre-existing model copies that are already adapted to different environments, simplifying the update process.
Data Source
AI summary
Provided is a model update method in a communication system, including: determining that a communication environment changes, and finding, according to a first mapping relationship set, a machine learning model suitable for a current communication environment from a plurality of machine learning models associated with the first mapping relationship set, to form a first machine learning model set; and deploying the machine learning model of the first machine learning model set on the access network device, and/or transmitting first information to a terminal, where the first information is used to instruct the terminal to deploy the machine learning model of the first machine learning model set.