Self-Feeding Deep Learning Edge Data Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional edge device data analysis methods are inefficient as they often fail to detect a large amount of data, leading to suboptimal run-time performance due to undetected results below a desired acceptance threshold.
Innovation Solution
A self-feeding deep learning method and system that re-trains edge-specific analytics and deploys an updated run model to analyze data, employing multiple thresholds to determine when to store and transfer data for further analysis, including an acceptance threshold, a consideration threshold, and an accumulation threshold to optimize data processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a single acceptance threshold is used to filter data analysis results, then the reliability of reported results is improved, but a large amount of potentially valuable data is lost
Solution Approach 1:
The patent segments the single acceptance threshold into multiple hierarchical thresholds: an acceptance threshold for high-confidence results, a consideration threshold for borderline cases, and a rejection threshold for clear failures. This segmentation allows the system to handle different confidence levels differently, preserving potentially valuable data while maintaining reliability for definitive results.
Solution Approach 2:
The consideration threshold acts as an intermediary between complete acceptance and complete rejection. Data results that fall between the acceptance and rejection thresholds are flagged for further review rather than being automatically discarded, serving as a buffer zone that prevents loss of potentially valuable borderline cases.
2Productivity
If all data below acceptance threshold is discarded, then the processing efficiency is improved, but the detection completeness deteriorates
Solution Approach 1:
The system dynamically adjusts the treatment of data based on its confidence level. High-confidence results above the acceptance threshold are immediately reported for efficient processing. Borderline results between thresholds are flagged for selective further analysis, and clear failures below the rejection threshold are discarded. This dynamic, multi-level approach optimizes both efficiency and completeness.
Solution Approach 2:
Different quality standards are applied to different data regions based on their confidence levels. Data above the acceptance threshold receives immediate reporting treatment, data in the consideration zone receives enhanced scrutiny, and data below the rejection threshold receives minimal processing. This local quality approach ensures appropriate resource allocation while maintaining detection completeness.
3Reliability
If multiple thresholds and re-training processes are implemented, then the data analysis completeness is improved, but the system complexity increases
Solution Approach 1:
The system performs preliminary filtering using the multi-threshold framework before initiating computationally intensive re-training processes. Only data in the consideration zone triggers re-training, while clear acceptances and rejections are handled immediately. This preliminary action prevents unnecessary complex processing and reduces overall system complexity.
Solution Approach 2:
The system uses its own operational data to automatically identify when re-training is needed, based on the accumulation of borderline cases. The multi-threshold framework enables the system to self-diagnose performance issues and trigger re-training only when necessary, reducing manual intervention and simplifying operation despite the underlying complexity.
Data Source
AI summary
Provided is a method and system for analyzing data in a run model at an edge device in a network environment. The method includes acquiring, at the edge device, data from the cloud environment, running a predetermined run model associated with the edge device and performing a first determination process by determining whether data analysis result from the run model performed is greater than an acceptance threshold. When it is determined that the data analysis result is less than the acceptance threshold, the method further performs a second determination process by determining whether the data analysis result is greater than a consideration threshold. If greater than the consideration threshold, the data is stored as acquired data to be further considered, and transferred to a cloud server.