Feature extraction method and system for high-dimensional heterogeneous industrial data
By performing feature grouping and global feature fusion on high-dimensional heterogeneous industrial data, the problems of existing technologies such as difficulty in adapting models to the distribution characteristics of heterogeneous data and lightweight design are solved, and efficient and stable feature extraction is achieved.
CN122432639APending Publication Date: 2026-07-21UNIV OF SCI & TECH BEIJING
View PDF 0 Cites 0 Cited by
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- UNIV OF SCI & TECH BEIJING
- Filing Date
- 2026-05-12
- Publication Date
- 2026-07-21
AI Technical Summary
Technical Problem
Existing technologies struggle to adapt to the differentiated distribution characteristics of different modalities and sources when processing high-dimensional heterogeneous industrial data. The models are complex and have a large number of parameters, which affects the efficiency of actual deployment.
Method used
By grouping industrial data by features, extracting features using independent branch networks, and combining a global feature fusion module and a multi-objective reconstruction loss function, the feature extraction process is optimized.
Benefits of technology
It improves the effectiveness of feature representation and the lightweight nature of the model, simplifies the computation and deployment complexity, and achieves efficient and stable feature extraction.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN122432639A_ABST
Abstract
The application provides a feature extraction method and system for high-dimensional heterogeneous industrial data, and relates to the technical field of data processing. The method comprises the following steps: obtaining high-dimensional heterogeneous industrial data; grouping the industrial data according to the feature attributes in the industrial data to obtain multiple feature subsets; inputting each feature subset into an independent branch network, performing feature extraction through the independent branch network, and obtaining multiple branch features; splicing each branch feature to obtain spliced features; inputting the spliced features into a global feature fusion module, performing global feature fusion through the global feature fusion module, and obtaining global fusion features; and combining a multi-target reconstruction loss function to constrain and optimize the global fusion features to obtain target extraction features.
Need to check novelty before this filing date? Find Prior Art