Characteristic selecting method and characteristic selecting device based on artificial neural network

A technology of artificial neural network and feature selection method, applied in the field of feature selection based on artificial neural network, can solve the problem of low efficiency of feature selection

Inactive Publication Date: 2016-07-20
NEC CORP
View PDF1 Cites 19 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] Since the importance measures of all features are calculated with the help of the trained artificial neural network, and then the features are sorted and selected, the feature selection efficiency of the above method is low

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
  • Characteristic selecting method and characteristic selecting device based on artificial neural network
  • Characteristic selecting method and characteristic selecting device based on artificial neural network
  • Characteristic selecting method and characteristic selecting device based on artificial neural network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0106] Figure 1a A schematic flowchart of a feature selection method based on an artificial neural network according to an embodiment of the present invention is shown. Such as Figure 1a As shown, the feature selection method based on artificial neural network may mainly include:

[0107] Step 101: Construct an artificial neural network with an input layer, an intermediate layer, and an output layer according to the K features to be selected and O output targets, wherein the input layer includes K nodes and each node corresponds to a feature, the The output layer includes O nodes and each node corresponds to an output target.

[0108] Step 102: Use the training set to train the artificial neural network to determine the connection weights from each layer to the next layer in the artificial neural network, wherein the optimization function used during training includes the sparsity of the input layer The item of constraint, so that the connection weight of the input layer to the ...

Embodiment 2

[0129] figure 2 A schematic flowchart of a feature selection method based on an artificial neural network according to another embodiment of the present invention is shown. figure 2 Win the mark and Figure 1a The same steps have the same functions. For brevity, detailed descriptions of these steps are omitted.

[0130] Such as figure 2 As shown, the difference from the foregoing embodiment is that after step 102 uses the training set to train the artificial neural network to determine the connection weights from each layer to the next layer in the artificial neural network, the artificial neural network is based on The feature selection method may further include: adding a test set including at least one test sample; according to the training set and the test set, using the trained artificial neural network to calculate the K features for the O output The impact of the target can specifically include:

[0131] Step 201: Calculate the first loss function represented by Formula 3...

Embodiment 3

[0152] Figure 3a A schematic flowchart of a feature selection method based on a deep learning network according to another embodiment of the present invention is shown. Figure 3a Win the mark and Figure 1a , figure 2 The same steps have the same functions. For brevity, detailed descriptions of these steps are omitted.

[0153] Such as Figure 3b As shown, a deep learning (Deep Learning with Feature Selection) network with feature selection can be constructed. The deep learning (Deep Learning) network is an artificial neural network with an intermediate layer (or hidden layer) greater than one. The deep learning network generally first initializes the parameters of each layer through unsupervised learning; then through supervised learning to fine-tune the overall parameters and optimize the parameters of all layers at the same time. In the embodiment of the present invention, feature selection (FeatureSelection) can be added to the input layer during unsupervised learning to in...

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 relates to a characteristic selecting method and a characteristic selecting device based on an artificial neural network. The characteristic selecting method is characterized in that the artificial neural network having the input layer, the middle layer, and the output layer can be formed according to K characteristics and O output targets; the artificial neural network can be trained by adopting a training set, and the connecting weight between every layer and the next layer of the artificial neural network can be determined; the optimization function used by the training comprises the items for the sparse constraint of the input layer, and then the connecting weight between the input layer and the next layer can be used to show the selecting result of the K characteristics. By adding the sparse constraint to the input layer of the artificial neural network, the training of the artificial neural network can be realized, and at the same time, the characteristic selecting result can be acquired, and then the efficiency of the characteristic selection of the artificial neural network can be improved.

Description

Technical field [0001] The present invention relates to feature selection in data mining, and in particular to a feature selection method and device based on artificial neural network. Background technique [0002] In the era of big data, there are many ways to collect data, so the characteristic dimension of the collected data is usually very large. However, when mining the required information from big data, not all features are beneficial to data mining. For example, some features are redundant, and some features may even hinder data mining. Therefore, feature selection is needed to remove redundant features and obstructive features in the data, thereby improving the efficiency of data mining and the effect of information extraction. [0003] At present, when data mining is performed, feature extraction can be performed based on a neural network (NeuralNetwork). For example, patent CN1945602A discloses a feature selection method based on neural network, which adds sparsity co...

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
Inventor 祁仲昂胡卫松
Owner NEC CORP
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