Training Data Collection Using Expert-Data Stop Criteria
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing learning models for inspection devices face challenges in achieving high classification accuracy due to insufficient evaluation of expert data quality and quantity, leading to increased generation costs and decreased accuracy when non-expert data is used.
Innovation Solution
A collecting device that includes units for acquiring and storing non-expert and expert defective product data, calculating standard deviations and rank sums of feature quantities, and determining the continuation and end of data acquisition based on standard deviation and rank sum ratios to optimize the collection of training data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If expert data is collected to improve classification accuracy, then learning model accuracy is improved, but collection cost increases
Solution Approach 1:
The patent changes the parameter of data quality evaluation from subjective expert judgment to objective statistical metrics (standard deviation and rank sum of feature quantities). By calculating these parameters for both non-expert and expert data, the system determines when sufficient expert data has been collected, preventing unnecessary data collection and associated costs while ensuring adequate classification accuracy.
2Measurement precision
If more expert data is collected, then learning model accuracy is improved, but generation cost increases
Solution Approach 1:
The patent implements a feedback mechanism where the system continuously calculates statistical parameters (standard deviation and rank sum) of collected expert data and compares them against corresponding parameters from non-expert data. When the ratio of these parameters meets predetermined thresholds, the system provides feedback to stop further expert data collection, thereby optimizing generation cost while maintaining adequate accuracy.
3Loss of energy
If non-expert data is used to reduce cost, then collection cost decreases, but classification accuracy decreases
Solution Approach 1:
The patent applies partial action by collecting only the necessary amount of expert data required to achieve sufficient classification accuracy, rather than collecting excessive expert data. The system determines this optimal point by comparing statistical parameters of expert and non-expert data, allowing cost-effective use of non-expert data while supplementing with minimal necessary expert data.
Data Source
AI summary
A first standard deviation α of feature quantities of all pieces of non-expert data stored in a non-expert data storage unit 13 is calculated, and a second standard deviation β of feature quantities of all pieces of expert data stored in an expert data storage unit 14 is calculated. In addition, a first rank sum ξ of the feature quantities of all pieces of non-expert data stored in the non-expert data storage unit 13 is calculated, and a second rank sum θ of the feature quantities of all pieces of expert data stored in the expert data storage unit 14 is calculated. Then, a continuation and an end of acquisition of defective product data by an expert are determined, based on the first standard deviation α and the second standard deviation β, and the first rank sum ξ and the second rank sum θ.