Edge Learning Data Selection with Ranking-Similarity Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing edge devices face constraints such as limited processing power, storage, connectivity, and energy, which hinder efficient data selection and storage for effective machine learning operations, necessitating an optimized data selection process to enhance learning quality and privacy.
Innovation Solution
An optimized data selection process involving a ranking-similarity-optimization (RSO) selection block that includes a ranking subsystem, similarity subsystem, and optimization subsystem to generate a list of data to be stored on the edge device, considering model relevance, data similarity, and device constraints.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all learning data is stored at the edge device, then learning quality is improved, but storage constraints are violated
Solution Approach 1:
The patent extracts only the most relevant and representative data samples from the complete learning dataset for storage at the edge device. By applying data selection algorithms that identify and extract high-value data points based on their contribution to model training, the system achieves effective learning with minimal storage requirements, resolving the contradiction between learning quality and storage capacity.
Solution Approach 2:
The patent applies local quality by differentiating between different types of data and assigning different storage priorities based on their specific characteristics and relevance to the machine learning model. Not all data is treated equally; instead, the system identifies and prioritizes storage of data with higher informational value, thereby achieving good learning quality with selective data storage rather than storing all data uniformly.
2Power
If data is transmitted to cloud for processing, then processing power is improved, but latency is increased
Solution Approach 1:
The patent segments the machine learning workload into two parts: data selection and preprocessing are performed at the edge device using limited local processing power, while only the selected critical data is transmitted to the cloud for heavy training operations. This segmentation allows immediate local processing to reduce latency while leveraging cloud power for computationally intensive tasks, resolving the contradiction between processing power and latency.
Solution Approach 2:
The patent performs preliminary data selection and filtering at the edge device before transmission to the cloud. By pre-processing the data locally to identify and select only the most relevant samples, the system reduces the amount of data that needs to be transmitted and processed in the cloud, thereby reducing overall latency while still utilizing cloud processing power effectively for the selected data.
3Measurement precision
If more data is stored at edge device, then learning accuracy is improved, but energy consumption is increased
Solution Approach 1:
The patent extracts only the essential and most informative data samples for storage at the edge device, rather than storing all available data. By applying selection criteria that identify data points with maximum informational value for model training, the system achieves high learning accuracy with a minimal subset of data, thereby reducing the energy required for data storage, management, and processing while maintaining model performance.
4Measurement precision
If data selection is performed manually, then selection precision is improved, but device complexity is increased
Solution Approach 1:
The patent implements self-service by enabling the edge device to automatically perform data selection using embedded algorithms that evaluate data relevance and importance. The system autonomously identifies and selects appropriate data samples without requiring manual intervention, thereby maintaining high selection precision through algorithmic evaluation while reducing operational complexity by eliminating manual data curation processes.
Data Source
AI summary
The present embodiments relate to systems, methods, and computer-readable media for selecting data to be stored at an edge device. More particularly, the present embodiments relate to an optimized data selection process for increased targeting of data observations that can be stored on the edge device for maximized edge learning quality. The selected data for storage at the edge can represent each class of a specified problem, maximizes the performance of the selected learning model, and can incrementally maintain a support set for further learning. The present embodiments can allow for robust learning on the edge with improved latency and improved privacy.