A structured multi-view hessian regularized sparse feature selection method
A sparse feature and feature selection technology, applied in instrument, computing, character and pattern recognition, etc., can solve the problems of not taking into account the characteristics of multi-view data, ignoring the correlation and complementary characteristics of different views, and achieve good feature selection. performance, the effect of improving performance
Active Publication Date: 2021-05-07
NORTH CHINA UNIVERSITY OF SCIENCE AND TECHNOLOGY
View PDF4 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
In recent years, semi-supervised feature selection methods based on Hessian regularization have been proposed. However, in the face of multi-view data, these semi-supervised feature selection methods have not considered the characteristics of multi-view data in the process of constructing Hessian regularization, ignoring Correlation and complementary characteristics between different view features
Method used
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0052] The structured multi-view Hessian regularization sparse feature selection method of the present invention will be described in detail below in conjunction with the accompanying drawings, including the following steps:
[0053] 1) Collect the underlying visual features of n original images to obtain m view image feature matrices, where,
[0054] The m view image feature matrices are:
[0055] X=(X v ) m×1 =[X 1 ,X 2 ,...,X m ] T ∈R d×n ,
[0056] In the formula (1), the d v is the feature dimension of the vth view image; the X v is the feature matrix of the vth view image, and In the formula (2), x 1 v ,x 2 v ...,x l v is the feature vector of the l labeled image under the vth view in the n original images, x l+1 v ,...,x n v Be the feature vector of n-l unlabeled images under the vth view in the n original images;
[0057] In the step 1), the underlying visual features include: color correlation map, wavelet texture and edge direction histogram, ...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More PUM
Login to View More Abstract
The invention discloses a structured multi-view Hessian regularization sparse feature selection method, comprising the following steps: collecting the underlying visual features of n original images, obtaining m view image feature matrices, and setting the feature selection mapping matrix of X as variable G , construct the objective function of structured multi-view Hessian regularization sparse feature selection, calculate the feature selection mapping matrix G of X through an iterative algorithm, and select the mapping matrix G according to the obtained features t , will be sorted in descending order, and the first ds features corresponding to X will be selected as the feature subset after feature selection. When performing semi-supervised feature selection on multi-view data, the present invention not only considers the importance of each view, but also considers the importance of different features under the same view. In addition, the use of multi-view Hessian regularization further improves the semi-supervised sparseness. Feature selection performance, therefore, the present invention has better feature selection performance.
Description
technical field [0001] The invention belongs to the technical field of semi-supervised sparse feature selection, and in particular relates to a structured multi-view Hessian regularized sparse feature selection method. Background technique [0002] In order to better understand, search and classify image data, many visual features have been proposed, such as shape features, color features, texture features, etc. Each type of feature describes the image data from a specific space, and has specific physical meaning and statistical characteristics. Traditionally, each type of feature can be regarded as a view, so data represented by different types of features is called multi-view data. How to obtain effective information of multi-view data has become a research hotspot in the field of feature selection analysis. [0003] One of the most direct methods is to directly concatenate multi-view data into a long feature vector. This method is simple, but this direct concatenation m...
Claims
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More Application Information
Patent Timeline
Login to View More Patent Type & Authority Patents(China)
IPC IPC(8): G06K9/46G06K9/48G06K9/62
CPCG06V10/507G06V10/513G06V10/478G06V10/56G06F18/2136
Inventor 史彩娟段昌钰赵丽莉刘利平葛超刘健闫晓东
Owner NORTH CHINA UNIVERSITY OF SCIENCE AND TECHNOLOGY