Pseudo Defective Data Generation Using Latent Variable Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional pseudo defective product data generation methods produce similar defective product data, limiting the improvement in inspection device determination accuracy due to the scarcity of actual defective product data.
Innovation Solution
A method involving a deep generation model, classification model, and distance learning model to generate pseudo defective product data by mixing features of non-defective and actual defective product data, creating latent variables and then generating additional pseudo defective product data that correlates with the actual data, thereby increasing the quantity of usable data for improved inspection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If conventional pseudo defective product data generation methods are used to create defective product data, then the quantity of defective product data increases, but the determination accuracy of inspection devices does not improve sufficiently because the generated data are similar to each other
Solution Approach 1:
The patent applies parameter changes by modifying the latent variables in the generative model through classification and distance learning processes. The latent variables are transformed to emphasize defective product characteristics while removing normal product characteristics, thereby changing the data distribution parameters to generate more diverse and accurate pseudo-defective data that improves inspection device determination accuracy
Solution Approach 2:
The patent uses latent variables as an intermediary between actual defective product data and pseudo-defective product data. The classification model and distance learning model process these latent variables to extract and emphasize defective product characteristics, creating a bridge that enables generation of high-quality pseudo data with improved determination accuracy
2Measurement precision
If many pieces of actual defective product data are collected to improve determination accuracy, then the accuracy improves, but it is difficult to collect sufficient defective product data since defective products are rare in manufacturing
Solution Approach 1:
The patent uses copying by generating pseudo-defective product data through a generative model that learns from actual defective product data. Instead of collecting numerous actual defective products, the system creates synthetic copies that preserve the essential defective characteristics, making the data collection process much easier while maintaining determination accuracy
Solution Approach 2:
The patent transforms the parameter distribution of latent variables through classification and distance learning to generate pseudo-defective data that captures the essence of actual defective products. This parameter transformation enables creation of sufficient training data without the need to collect rare actual defective products
Data Source
AI summary
Preparing actual defective product data and non-defective product data, causing a deep generation model to learn the data and to generate latent variables in which features of the non-defective product data and the actual defective product data are mixed, causing a classification model to learn the latent variables to generate a classified non-defective product and defective product latent variable, deleting the non-defective product latent variable from the classified non-defective product and defective product latent variable to output the defective product latent variable including a gray latent variable, causing a distance learning model to learn the defective product latent variable and the non-defective product latent variable and to delete the gray latent variable, and causing the deep generation model to learn the defective product latent variable that has been obtained and to generate the pseudo defective product data greater in number than the actual defective product data.