AI/ML Model Training via Terminal Measurement Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing communication technologies face challenges in properly training artificial intelligent/machine learning (AI/ML) models to ensure satisfying communication performances.
Innovation Solution
A method for communication that involves receiving a configuration indicating a subset of resources at a terminal device or network device for constructing a dataset to train an AI/ML model, and transmitting information on quality determination based on these resources to optimize the model training.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If AI/ML models are trained using existing communication technologies, then communication performance can be improved, but the model training accuracy and reliability are insufficient
Solution Approach 1:
The patent implements a feedback mechanism where the terminal device reports measurement results (including quality determination information) back to the network device. This feedback loop enables the network device to construct accurate datasets based on real measurement results, thereby improving model training accuracy and reliability. The feedback process involves: terminal device performing measurements, evaluating quality, and reporting results to the network device for dataset construction.
Solution Approach 2:
The patent applies preliminary action by having the terminal device perform measurements and quality evaluations before the network device constructs the training dataset. The terminal device preliminarily processes the measurement data by determining quality metrics and selecting relevant information, which then forms the basis for accurate model training. This preliminary processing ensures that only high-quality, relevant data is used for training.
2Measurement precision
If comprehensive datasets are constructed for AI/ML model training, then model training accuracy improves, but the complexity of data collection and processing increases
Solution Approach 1:
The patent segments the data collection process by dividing the dataset into different types (training dataset, validation dataset, test dataset) and having the terminal device selectively report only the necessary measurement results for each purpose. The network device then segments the processing by constructing different datasets from the reported information based on specific training needs. This segmentation reduces complexity by avoiding unnecessary data collection and processing.
Solution Approach 2:
The patent extracts only the essential information needed for model training from the comprehensive measurement results. The terminal device extracts quality determination information and selects relevant measurement data, while the network device extracts and processes only the necessary portions for constructing training, validation, and test datasets. This extraction process reduces data collection and processing complexity while maintaining training accuracy.
3Reliability
If real measurement results are used for dataset construction, then communication performance improves, but the time required for measurement and data processing increases
Solution Approach 1:
The patent applies preliminary action by having the terminal device perform measurements and quality evaluations in advance before the network device constructs the training dataset. The terminal device preliminarily processes the measurement data by determining quality metrics and preparing relevant information, which reduces the overall time required for data collection and processing. This preliminary processing allows the network device to receive ready-to-use data for immediate dataset construction.
Solution Approach 2:
The patent ensures continuity of useful action by implementing an ongoing process where the terminal device continuously performs measurements and reports results to the network device. This continuous feedback loop enables real-time dataset construction and model training without interruption. The continuous process eliminates idle time and ensures that measurement and processing activities are always productive, reducing overall time loss.
Data Source
AI summary
Embodiments of the present disclosure relate to methods, devices, and computer readable medium for communication. According to embodiments of the present disclosure, a terminal device receives a measurement configuration or transmission configuration from a network device. The terminal device performs a measurement based on the measurement configuration and report results of the measurement to the network device. Alternatively, if the terminal device performs transmissions based on the transmission configuration, the network device performs a measurement based on the transmissions. The network device trains an intelligent (AI) or machine learning (ML) based on the results of the measurement. In this way, the real measurement results are used for constructing a suitable dataset for the AI or ML model at the network device.