Generation Model Training Using Federated Identification Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods face challenges in constructing a generation model that can generate a wide variety of data due to limitations in collecting sufficient and diverse learning data across different sites, leading to inaccurate visual inspection and inference, while also posing issues with data confidentiality and high communication and calculation costs.
Innovation Solution
A model generation apparatus that generates data using a generation model, transmits it to identification models for identification, collects identification results, and trains the model to degrade specific identification performances, ensuring low communication and calculation costs and maintaining data confidentiality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a sufficient amount of inference learning data is collected to improve visual inspection accuracy, then the accuracy of visual inspection is improved, but the cost of data collection increases
Solution Approach 1:
The patent uses a generation model to create synthetic inference learning data that copies the statistical distribution characteristics of real learning data. Instead of collecting additional real images, the system generates artificial images that replicate the essential features and variations present in real product images, thereby increasing data quantity without proportionally increasing collection costs
Solution Approach 2:
The generation model transforms the approach from direct data collection to parameter-based data synthesis. By learning the distribution parameters from a small set of real images and generating new samples based on these parameters, the system can produce large quantities of training data with controlled variations, reducing the need for extensive physical data collection
2Adaptability or versatility
If learning data is gathered from multiple sites to increase data variety, then the variety of learning data is improved, but data confidentiality and communication costs increase
Solution Approach 1:
The patent introduces a generation model as an intermediary that receives distribution information from multiple sites without directly handling the actual learning data. Each site can contribute to training the generation model or validate its output without sharing their proprietary real images, thus maintaining data confidentiality while still achieving diverse data generation through the intermediary model
Solution Approach 2:
Instead of gathering and sharing actual learning data from multiple sites, the system creates copies of the statistical distribution characteristics from each site's data. The generation model learns to reproduce the essential features and variations from multiple sources without requiring direct access to or transmission of the original confidential images
Data Source
AI summary
A model generation apparatus according to one or more embodiments may include: a generating unit that generates data using a generation model; a transmitting unit that transmits the generated data to a plurality of trained identification models that each have acquired, by machine learning using local learning data, a capability of identifying whether given data is the local learning data, and causes the identification models to perform an identification on the data; a receiving unit that receives results of identification with respect to the transmitted data executed by the identification models; and a learning processing unit that trains the generation model to generate data that causes identification performance of at least one of the plurality of identification models to be degraded, by performing machine learning using the received results of identification.