Continual Learning Defect Diagnosis With Influence-Based Buffer Updates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing AI-based radiographic testing methods face challenges in maintaining accuracy over time due to catastrophic forgetting, which occurs when relearning all previous data sets, leading to storage space issues.
Innovation Solution
A method and device for defect diagnosis using continual learning, involving training a current model with a buffer training data set and a current data set, calculating influence and change degrees, and updating the buffer data set to retain knowledge while adapting to new data without excessive storage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all previous training data sets are relearned to prevent catastrophic forgetting, then accuracy of previous data is maintained, but storage space requirements increase excessively
Solution Approach 1:
The patent segments the training data into a buffer data set (containing selected previous training data) and current training data set. Instead of storing and relearning all previous data, only a segmented portion is retained in the buffer, reducing storage requirements while maintaining accuracy through selective data retention.
Solution Approach 2:
The patent extracts and stores only the most influential previous training data points in a buffer, rather than retaining all previous data. This extraction approach identifies and keeps only the critical data needed to prevent catastrophic forgetting, significantly reducing storage space while maintaining model performance.
2Quantity of substance
If a buffer training data set is used to reduce storage space, then storage requirements decrease, but catastrophic forgetting may occur reducing accuracy
Solution Approach 1:
The patent applies local quality by selectively choosing which previous training data points to store in the buffer based on their influence degree on model accuracy. Not all previous data is treated equally; instead, only data points with high influence are retained, optimizing the buffer content to prevent catastrophic forgetting while minimizing storage requirements.
Solution Approach 2:
The patent changes the parameter of data selection from random or uniform sampling to influence-based sampling. By calculating and using influence degrees of previous data points, the system dynamically adjusts which data to retain in the buffer, ensuring that the most critical data is preserved to prevent catastrophic forgetting.
3Reliability
If continual learning is implemented with selective data updating, then model accuracy is maintained over time, but computational complexity increases
Solution Approach 1:
The patent performs preliminary action by pre-calculating and storing the influence degrees of previous training data points before the actual training process. This preliminary computation allows the system to quickly select appropriate buffer data during continual learning without adding significant computational overhead during the main training phases, thus maintaining model accuracy while managing complexity.
Data Source
AI summary
Proposed are a device for diagnosing a defect using a model based on continual learning and a method therefor, and the method for diagnosing the defect includes a step of loading, by a training unit, a buffer training data set, which is training data selected according to a degree of influence on prediction performance of a current model from among a past training data set in the continual learning, a step of training the current model using the buffer training data set and a current training data set, and a step of detecting, by a detection unit, the defect in a radiographic image using the current model when the radiographic image is input.