ML Model Fine-Tuning Across Network Devices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
There is no mechanism to reuse or further fine-tune machine learning models trained or fine-tuned using online training for a single UE by other UEs in wireless communication systems, limiting the utilization of updated models across multiple network devices.
Innovation Solution
The system allows for the fine-tuning of machine learning models across multiple UEs by transmitting configuration information and parameters, enabling other UEs with similar hardware to use and further fine-tune the models, and sharing the fine-tuned models across a network entity and other UEs, using techniques such as federated learning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If online training is used to fine-tune an ML model for a single UE, then the model can be adapted to specific UE conditions, but there is no mechanism to reuse or further fine-tune the model by other UEs
Solution Approach 1:
The patent combines online training capabilities across multiple UEs by enabling a first UE to fine-tune a model and then share the fine-tuned model with second and third UEs. This merging approach allows individual UE adaptations to be consolidated and redistributed, resolving the contradiction between model adaptability and reuse efficiency.
Solution Approach 2:
The patent makes the ML model universal by enabling it to be fine-tuned for one UE and then reused by multiple other UEs with similar hardware. This multi-functionality allows a single model to serve multiple purposes and multiple devices, simultaneously achieving adaptability and reuse efficiency.
2Productivity
If offline training is performed to fine-tune an ML model, then the fine-tuned model can be easily deployed across multiple UEs, but the model cannot be dynamically updated based on real-time UE-specific data
Solution Approach 1:
The patent applies preliminary action by performing offline training to create an initial ML model that can be easily deployed across multiple UEs. This pre-trained model serves as a foundation that can then be further adapted through online training, resolving the contradiction between deployment efficiency and real-time adaptation.
Solution Approach 2:
The patent introduces dynamics by enabling the initially offline-trained model to be subsequently fine-tuned through online training mechanisms. This allows the model to transition from a static, pre-trained state to a dynamic state where it can be continuously adapted based on real-time UE-specific data while maintaining easy deployability.
Data Source
AI summary
An apparatus, method and computer-readable media are disclosed for performing wireless communications. For example, a first network device can transmit, to one or more second network devices, configuration information associated with a trained machine learning model. The first network device can receive, from the one or more second network devices, information associated with a first fine-tuned machine learning model based on adaptation of parameters of the trained machine learning model. The first network device can further output, for transmission to one or more third network devices, configuration information associated with the first fine-tuned machine learning model.