ANN Object Classification Using Orthogonal Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional classification techniques in AI systems are limited in accuracy and efficiency due to insufficient training data and incomplete definition of classes, failing to effectively identify and recognize object-level components in a contextual manner, and are not sufficient for content classification.
Innovation Solution
The method involves extracting positive and orthogonal features from input data using an artificial neural network (ANN) model, where positive features contribute to identifying a class, and orthogonal features contribute to identifying other classes, with a first part of the ANN model performing partial classification based on pattern detection and a second part determining classification accuracy by detecting the absence of patterns.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional classification techniques are used, then the system is simpler to implement, but classification accuracy and efficiency are limited due to insufficient training data and incomplete class definitions
Solution Approach 1:
The classification system is segmented into two distinct parts: a first part that performs partial classification using positive features (patterns that define a class), and a second part that determines classification accuracy using orthogonal features (patterns that distinguish between classes). This segmentation allows each part to specialize in specific aspects of classification, improving overall accuracy without requiring a monolithic complex system.
Solution Approach 2:
The invention introduces a new dimension to classification by incorporating orthogonal features that detect the absence of patterns. Instead of only analyzing what defines a class (positive features), the system also analyzes what distinguishes it from other classes (orthogonal features). This dimensional expansion from one-sided pattern detection to two-sided pattern analysis significantly improves classification accuracy.
2Productivity
If conventional pattern detection is used, then the method is simpler, but it fails to efficiently identify and recognize object-level components in a contextual manner
Solution Approach 1:
Object recognition is segmented into two functional parts: the first part identifies object-level components using positive features, while the second part recognizes contextual relationships using orthogonal features. This segmentation enables efficient processing by dividing the complex task of contextual object recognition into specialized sub-tasks.
Solution Approach 2:
The system transitions from single-dimensional pattern detection to two-dimensional analysis by incorporating orthogonal features that capture contextual relationships between objects. This dimensional enhancement allows efficient object-level component identification while simultaneously understanding contextual relationships without requiring overly complex processing.
3Measurement precision
If incomplete class definitions are used, then the system is easier to train, but identification and recognition are impacted by incomplete definition of class
Solution Approach 1:
The system addresses incomplete class definitions by adding a second dimension of analysis through orthogonal features. These features capture patterns that are present in one class but absent in others, providing complementary information that completes the class definition without requiring additional training data for each individual class.
Solution Approach 2:
The orthogonal features serve multiple functions: they define class boundaries, distinguish between classes, and provide contextual relationships. This multi-functionality reduces the need for extensive training data by making the feature set more informative and versatile.
Data Source
AI summary
This disclosure relates to method and system for classifying an object in input data using an artificial neural network (ANN) model. The method may include extracting positive features and orthogonal features associated with the object in the input data, performing a partial classification of the object based on the positive features by a first part of the ANN model, and determining an accuracy of the classification of the object based on the orthogonal features by a second part of the ANN model. The positive features are features uniquely contributing to identification of a class for the object, while the orthogonal features are features not contributing to identification of the class but contributing to identification of one or more of remaining classes.