Wireless UE Data Collection With Similarity-Based Sample Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing wireless communications systems face challenges in balancing the collection of sufficient data samples for AI/ML model training and monitoring while minimizing overhead and maintaining performance and accuracy.
Innovation Solution
Implementing a reduced sample size mode of operation for data collection, where UEs determine and apply a reduced set of samples based on control information from network entities, using similarity metrics and other conditions to select data for logging and reporting.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a reduced sample size mode of operation is implemented for data collection, then overhead is minimized and data collection efficiency is improved, but the quantity of collected data samples is reduced
Solution Approach 1:
The patent extracts only the most relevant and informative data samples from the complete dataset using similarity metrics and selection criteria. Instead of collecting all available data samples, the system identifies and extracts a reduced subset that maintains representativeness and diversity, thereby reducing overhead while preserving data quality for AI/ML model training.
Solution Approach 2:
The patent changes the parameter of sample size from the traditional large-scale collection to a reduced sample size mode. By adjusting this parameter and applying similarity-based selection criteria, the system transforms the data collection process to collect fewer but more representative samples, improving efficiency while maintaining model training effectiveness.
2Quantity of substance
If all measured data samples are collected and reported, then the quantity of training data is maximized, but overhead and resource consumption increase
Solution Approach 1:
The system extracts a representative subset of data samples from the complete measured dataset by applying similarity metrics and selection criteria. This extraction process identifies and retains only the most valuable samples for training, discarding redundant ones, thereby reducing the quantity of data that needs to be transmitted and processed while maintaining training effectiveness.
Solution Approach 2:
Instead of collecting and reporting all measured data samples (excessive action), the patent applies partial action by collecting and reporting only a reduced subset of samples that are sufficient for effective model training. This partial collection approach reduces overhead and resource consumption while maintaining the essential quality and representativeness of the training data.
3Productivity
If a reduced set of samples is used for data collection, then overhead is reduced and efficiency is improved, but measurement precision may be degraded
Solution Approach 1:
The patent applies local quality by ensuring that the reduced sample set maintains high quality characteristics in critical dimensions. Using similarity metrics and diversity criteria, the system identifies and selects samples that locally represent different regions and characteristics of the complete dataset, thereby preserving overall data representativeness and measurement precision despite the reduced sample size.
Solution Approach 2:
The patent changes the selection parameters from random or uniform sampling to similarity-based and diversity-aware sampling. By adjusting these parameters and applying metrics that evaluate representativeness and coverage, the system transforms the reduced sample set into one that maintains high measurement precision and data quality, counteracting the potential degradation from reduced sample size.
Data Source
AI summary
Various aspects of the present disclosure relate to data collection procedures that reduce the number of samples reported to a network entity and/or logged by a user equipment (UE), without introducing degradation in performance and/or accuracy of a trained artificial intelligence/machine learning (AI/ML) model or a related monitoring scheme. For example, the UE may receive control information that indicates a reduced sample size mode of operation associated with data collection, determine whether to apply the reduced sample size mode of operation for a measured set of data, and incorporate the measured set of data into a reduced set of samples based on the determination.