Neural Network Classification Device Heterogeneous Data Integration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current machine learning technologies are unable to effectively train or improve classification models using heterogeneous sensing data from IoT devices, limiting the accuracy of classification results.
Innovation Solution
A classification device and method based on a neural network that incorporates a heterogeneous integration module with a convolutional layer, data normalization layer, connected layer, and classification layer, coupled with a recurrent neural network, which generates classification results from heterogeneous data, including image and numerical data across different time points.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If current machine learning technology is used to train classification models, then the model can process single-type data, but it cannot effectively utilize heterogeneous sensing data from IoT devices
Solution Approach 1:
The patent merges multiple types of neural network layers (convolutional layers for image data, fully connected layers for numerical data) into a single heterogeneous integration module. This allows the system to process both image data and numerical sensing data simultaneously through unified feature extraction and classification, enabling effective utilization of heterogeneous IoT data while maintaining high classification accuracy.
Solution Approach 2:
The heterogeneous integration module is designed with multi-functionality to handle different data types. The convolutional layers process image data while fully connected layers process numerical data, and both pathways converge in the classification layer. This universal design enables the single module to adapt to various IoT data sources and types without requiring separate processing systems.
2Quantity of substance
If heterogeneous data from multiple time points is collected, then more training data is available, but the model cannot effectively improve classification accuracy through temporal data
Solution Approach 1:
The patent implements continuous processing of heterogeneous data streams from multiple time points through the heterogeneous integration module. The module continuously extracts features from incoming image and numerical data, normalizes them, and feeds them through the classification layers in real-time. This continuous action allows the model to accumulate and utilize temporal patterns from multiple time points, improving classification accuracy through the sequential processing of time-series IoT data.
3Ease of manufacture
If a simple neural network structure is used, then the model is easier to train, but it cannot effectively process multiple data types simultaneously
Solution Approach 1:
The patent segments the neural network into distinct functional components: convolutional layers for image data processing, fully connected layers for numerical data processing, and a unified classification layer. Each segment is specialized for specific data types but works together in a coordinated manner. This segmentation allows each component to be trained independently on its specific data type while maintaining overall system versatility for processing multiple heterogeneous data types simultaneously.
Data Source
AI summary
A classification device and a classification method based on a neural network are provided. A heterogeneous integration module includes a convolutional layer, a data normalization layer, a connected layer and a classification layer. The convolutional layer generates a first feature map according to a first image data. The data normalization layer normalizes a first numerical data to generate a first normalized numerical data. The first numerical data corresponds to the first image data. The connected layer generates a first feature vector according to the first feature map and the first normalized numerical data. The classification layer generates a first classification result corresponding to a first time point according to the first feature vector. The heterogeneous integration module generates a second classification result corresponding to a second time point. A recurrent neural network generates a third classification result according to the first classification result and the second classification result.