Machine Learning Classifier Scene Graph Analysis for Misclassification Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning classifiers often make systematic mistakes in classifying digital images, which can lead to misclassifications and anomalies that are not detected.
Innovation Solution
A device and computer-implemented method that provide a rule to identify systematic mistakes in classifiers by analyzing scene graphs of digital images, distinguishing between correctly and incorrectly classified images, and determining a logic program to identify patterns leading to misclassifications.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a classifier is used to classify digital images, then classification capability is provided, but systematic mistakes and misclassifications occur
Solution Approach 1:
The patent implements a feedback mechanism where the system analyzes misclassified images, determines scene graphs, identifies systematic mistakes through logic programs, and uses this information to improve the classifier. The feedback loop includes: classifying images, detecting misclassifications, analyzing scene graphs, determining rules for systematic mistakes, and updating the classifier accordingly.
Solution Approach 2:
The system performs preliminary analysis by pre-determining scene graphs for images before final classification, and by pre-identifying systematic mistake patterns through logic programs. This allows the system to anticipate and prevent misclassifications before they occur in production.
2Reliability
If scene graphs are analyzed to identify systematic mistakes, then anomaly detection capability is improved, but device complexity increases
Solution Approach 1:
The patent segments the analysis process into distinct components: image classification module, scene graph generation module, misclassification detection module, logic program determination module, and classifier update module. Each component handles a specific aspect of the analysis, making the overall complex system more manageable and maintainable.
Solution Approach 2:
The scene graph serves as an intermediary representation that bridges the gap between raw image data and classification decisions. By introducing this intermediate structure, the system can analyze relationships and patterns without directly manipulating complex pixel data, simplifying the anomaly detection process.
3Reliability
If the classifier is improved through rule-based learning, then classification accuracy improves, but training time and computational resources increase
Solution Approach 1:
The system changes the parameter representation from raw pixel values to high-level semantic representations (scene graphs and logic programs). This transformation allows the training process to work with more meaningful features that capture essential patterns, reducing the computational burden and training time while improving accuracy.
Data Source
AI summary
A device and computer-implemented method for machine learning. The method includes: providing first and second classes; providing a first set of scene graphs including scene graphs of digital images that are incorrectly classified in the first class or the second class; providing a second set of scene graphs including scene graphs of digital images that are correctly classified with respect to the first class or the second class; determining, depending on the first set of scene graphs and the second set of scene graphs a rule that indicates that a presence of a first object and/or a second object in a digital image and/or a relation between the first object and the second object in the scene graph of the digital image results in that the classification of the digital image includes a misclassification of the digital image into the second class instead of the first class.