Visualized Feature Vector Image Classification via Deep Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current machine learning methods, such as Support Vector Machines (SVM), face challenges in nonlinear classification and scalability for multi-class classification tasks, requiring kernel function selection and increased computational complexity as the number of classes grows, limiting the use of deep-learning AI networks like CNNs for data classification based on pre-selected feature vectors.
Innovation Solution
An artificial intelligence neural network apparatus and method that visualizes feature vectors into image databases, using deep-learning techniques to perform image classification, incorporating a feature vector to image conversion unit that synthesizes cross-correlation images and local pattern images, enabling efficient classification with a deep-learned neural network.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If SVM is used for nonlinear classification by mapping data to high-dimensional feature vector spaces, then classification capability is improved, but device complexity increases due to kernel function selection and computation requirements
Solution Approach 1:
The patent replaces the mechanical SVM classification system with a deep learning neural network system that automatically learns feature representations. The neural network substitutes the manual kernel function selection process with automated feature learning through multiple hidden layers, eliminating the need for explicit kernel function choices while achieving superior nonlinear classification performance
Solution Approach 2:
The patent transforms the classification problem from traditional feature vector space to image space by converting feature vectors into visual representations. This dimensional transformation allows the use of convolutional neural networks that operate on image data, enabling the system to leverage spatial hierarchies and local patterns in the transformed feature space for improved classification
2Adaptability or versatility
If multiple SVMs are coupled to achieve multi-class classification, then classification coverage is improved, but productivity decreases due to exponential increase in computation amount
Solution Approach 1:
The patent merges multiple classification functions into a single deep learning neural network model. Instead of coupling multiple SVMs to handle different class combinations, the neural network processes all classes simultaneously through a unified architecture with shared feature extraction layers, dramatically reducing computational overhead while maintaining comprehensive multi-class classification capability
Solution Approach 2:
The patent performs preliminary feature extraction and representation learning in the neural network's hidden layers before final classification. This preliminary action of learning robust feature representations upfront enables the model to efficiently handle multi-class classification without requiring repeated computations for different class pairs, improving overall productivity
3Measurement precision
If machine learning is used for data classification based on feature vectors, then classification accuracy is improved, but adaptability decreases because deep-learning networks cannot be used
Solution Approach 1:
The patent introduces an intermediary component that converts traditional feature vectors into image representations. This intermediary transformation layer acts as a bridge between conventional machine learning feature extraction and deep learning image processing networks, enabling CNNs and other image-based deep learning models to process structured feature data while maintaining the accuracy benefits of feature-based classification
Solution Approach 2:
The patent changes the parameter representation from traditional feature vector format to image matrix format. By transforming the data structure and representation parameters, the system enables compatibility with deep learning networks that expect image inputs, while preserving the discriminative power of hand-crafted or pre-selected feature vectors through the transformation process
Data Source
AI summary
An artificial intelligence neural network apparatus, comprising: a labeled learning database having data of a feature vector composed of N elements; a first feature vector image converter configured to visualize the data in the learning database to form an imaged learning feature vector image database; a deep-learned artificial intelligence neural network configured to use a learning feature vector image in the learning feature vector image database to perform an image classification operation; an inputter configured to receive a test image, and generate test data based on the feature vector; and a second feature vector image converter configured to visualize the test data and convert the visualized test data into a test feature vector image. The deep-learned artificial intelligence neural network is configured to determine a class of the test feature vector image.