Small sample multi-classification method based on nested integrated depth support vector machine

A support vector machine, multi-classification technology, applied in neural learning methods, computer parts, character and pattern recognition, etc., can solve problems such as high computational cost, deviation of sample distribution estimates, large amounts of data, etc., to improve classification accuracy , in exchange for the flexible effect of the kernel function

Pending Publication Date: 2020-04-10
TIANJIN NORMAL UNIVERSITY
View PDF6 Cites 4 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, without considering the training sample tasks and data properties, improving the data plane often leads to biased sample distribution estimates, which may affect the realism of the classification
Improved generative models trained with prior knowledge require large amounts of data and high computational cost
Meanwhile, meta-learning methods cannot avoid polluting learners with irrelevant tasks

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 more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Small sample multi-classification method based on nested integrated depth support vector machine
  • Small sample multi-classification method based on nested integrated depth support vector machine
  • Small sample multi-classification method based on nested integrated depth support vector machine

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0066] A small-sample multi-classification method based on nested ensemble deep support vector machine, such as figure 1 The structural diagram of the nested ensemble DSVM algorithm is shown. It is a deep support vector machine (Deep Support Vector Machine, DSVM) algorithm with three different single-kernel functions as a sub-classifier to build a base classifier; it will use a "one-to-one" strategy to convert the multi-category classification problem into multiple binary Classification problems; through the AdaBoost framework to complete multi-classification tasks, including:

[0067] (1) Sub-classifier model

[0068] figure 1 The principle of two-layer DSVM is described, and the input with label training data for ,

[0069] ; where, are called training sample points; these training samples, after containing in a DSVM core layer To extract the corresponding features, given to ;Finally, by a main SVM named M, the above eigenvalues ​​are processed and predicted...

Embodiment 2

[0112] We choose a separate SVM and endow it with the same kernel functions as the subclassifier DSVM, namely radial basis kernel function, Sigmod kernel function and polynomial kernel function; the same training data and test data are used to select the SVM The kernel function is trained and tested to obtain the correlation coefficient and hyperplane equation expression that needs to be obtained; calculate the geometric mean distance from all test samples to the hyperplane according to the obtained parameter values; use all the correct test samples to obtain The reciprocal of the mean square error of the distance between the distance and the geometric mean distance of all test samples is used as the weight value corresponding to the kernel function, and this is given to the base classifiers DSVM1, DSVM2, and DSVM3; Table 1:

[0113]

[0114] From Table 1, it can be seen that each sub-classifier The weight value after selecting the kernel function . Among them, the radi...

Embodiment 3

[0116] We built a system for detecting human objects through walls. The experimental equipment used in this system is a P410MRM single base station radar module with small size, low power consumption and simple operation. It is equipped with antenna ports for dual antenna operation. In experiments, the system uses this module to detect human targets behind walls and obtain reflected human status signals. Meanwhile, three indoor experimental schemes were designed. Including the state of no one behind the wall, the state of one person breathing rapidly behind the wall, and the state of two people walking slowly at 0.1m / s behind the wall. The wall involved in the experiment is a brick wall. The thickness of the brick wall is 25 cm. The distance between the human target and the wall is 100 cm. The distance between the radar equipment and the wall is 60 cm, and the human target is separated from the brick wall. On the basis of the above-mentioned experimental system, we colle...

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

No PUM Login to view more

Abstract

The invention discloses a small sample multi-classification method based on a nested integrated depth support vector machine. According to the method, a two-layer integrated classification technologyis adopted to analyze multi-class classification tasks under the condition of small samples; for an internal layer, three different single-kernel functions (a radial basis kernel function, a Sigmod kernel function and a polynomial kernel function) are respectively given to a deep support vector machine (DSVM) of a sub-classifier, and a basis classifier is constructed through integrated classification; for an external layer, a'one-to-other 'strategy is used to convert a multi-class classification problem into a plurality of dichotomy problems. And finally, an AdaBoost framework is utilized to complete a multi-classification task. The method can effectively avoid an'over-fitting 'phenomenon under the condition of small samples, so that the accuracy of through-wall human body target recognition under the condition of small samples is improved.

Description

[0001] This application was supported by the National Natural Science Foundation of China projects No.61501326 and No.61971310. technical field [0002] The invention relates to a technology of data mining under the condition of small samples. The method combines the nested integrated classification technology idea, deep learning theory and the advantages of support vector machines in eliminating over-fitting, and realizes multi-category under the condition of small samples. Classification. Background technique [0003] With the rapid development of artificial intelligence technology, deep learning algorithms have achieved remarkable results in many fields such as computer vision, natural language processing, speech recognition, image classification and detection. However, the success of these cases often relies on the collection of large amounts of expensive labeled data. This has certain difficulties in practical application scenarios. Furthermore, how to effectively lea...

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
no application Login to view more
Patent Type & Authority Applications(China)
IPC IPC(8): G06K9/62G06N3/08
CPCG06N3/084G06F18/2411G06F18/2431
Inventor 王为张丽王志雄张梦君
Owner TIANJIN NORMAL UNIVERSITY
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap