Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Feature selection based multi-example multi-tag learning method and system

A technology of feature selection and learning method, applied in the field of pattern recognition and machine learning, can solve the problems of being susceptible to noise, loss of label related information, sensitive to the number of classifier clusters, etc., to improve the labeling effect, increase the available information, The effect of improving accuracy

Inactive Publication Date: 2015-11-11
LUDONG UNIVERSITY
View PDF2 Cites 13 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Currently, there are two methods to solve this problem, one can degenerate the multi-instance multi-label problem into a multi-instance single-label problem, and use the multi-instance single-label method based on example selection to solve the multi-instance multi-label problem, but this type of method The predictions of each label in are independent of each other, which will lead to the loss of label-related information
Another method is to select examples directly based on multi-instance and multi-label, and then perform multi-label simultaneous prediction. The existing method is the KISAR algorithm, which can better consider the interaction between labels while selecting key examples. relationship, however, the selection of key examples in the KISAR algorithm uses a clustering method, which is easily affected by noise, and the cluster center as a key example lacks pertinence. unstable

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
  • Feature selection based multi-example multi-tag learning method and system
  • Feature selection based multi-example multi-tag learning method and system
  • Feature selection based multi-example multi-tag learning method and system

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0025] Embodiment 1. A multi-instance multi-label learning method based on feature selection. Combine below figure 1 The method provided in this embodiment will be described in detail.

[0026] see figure 1 , S1. Map all package features in the multi-instance and multi-label known data set to the projected example reference space composed of all examples in the known data set, and obtain the feature vector of the package.

[0027] Specifically, the multi-instance and multi-label known data set specifically includes: the number of packages and packages in the known data set, the known examples and the number of examples in each package, the known feature vector representing each example, The label set corresponding to each package and the label set corresponding to all packages are known, wherein the known multi-instance multi-label data set is specifically expressed as {(X 1 ,Y 1 ),(X 2 ,Y 2 ),…,(X i ,Y i ),…,(X m ,Y m )},in, Package X for known datasets i n in i...

Embodiment 2

[0101] Embodiment 2. A method of predicting the label set of multi-instance and multi-label unknown samples using a multi-instance and multi-label learning method based on feature selection. Combine below figure 2 The method provided in this embodiment will be described in detail.

[0102] see figure 2 , this embodiment provides a method for predicting the label set of multi-instance and multi-label unknown samples by using a multi-instance multi-label learning method based on feature selection.

[0103] S51. Map all unknown package features in the multi-instance and multi-label unknown samples to be predicted to the representative projection example set In the reference space α of the representative projection example composed of , and then obtain the feature set of all unknown packages, where the unknown package P i After feature mapping, the q-dimensional feature vector ψ(P i ).

[0104] Specifically, the unknown packet P in the multi-instance multi-label unknown sa...

Embodiment 3

[0118] Embodiment 3, a multi-instance multi-label learning system based on feature selection. Combine below image 3 The system provided in this embodiment will be described in detail.

[0119] see image 3 A multi-instance multi-label learning system based on feature selection provided in this embodiment includes a known data set storage unit 61, a first feature mapping unit 62, a feature selection unit 63, a second feature mapping unit 64, and label correlation classification The feature selection unit 65 specifically includes an objective function obtaining module 631, an objective function solving module 632, a projected example removal module 633 and a reference space obtaining module 634, and the label correlation classifier unit 65 specifically includes a linear decision function module 651 , a correlation matrix definition module 652 , a classifier objective function module 653 and a label correlation classifier obtaining module 654 .

[0120] Among them, the known ...

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 present invention discloses a feature selection based multi-example multi-tag learning method and system. The method comprises: mapping all features of a packet in a known data set into a projection example reference space consisting of all examples in the known data set, and obtaining a feature vector of the packet; removing projection examples corresponding to the features of the packet with an invalidly marked tag in the projection example reference space by adopting a feature selection method based on l2 and 1 norm constraints, and further obtaining a representative projection example reference space; mapping the features of the packet into the representative projection example reference space, and obtaining a new feature vector of the packet; and constructing a linear decision function according to the new feature vector of the packet, and training a classifier based on tag correlation by adopting an optimal algorithm based on the tag correlation. By using the method provided by the present invention, the classifier based on the tag correlation is obtained by learning by utilizing the known database, so that a tag set of unknown samples is predicted and the accuracy of tag prediction is improved.

Description

technical field [0001] The invention relates to the field of pattern recognition and machine learning, in particular to a multi-instance multi-label learning method and system based on feature selection. Background technique [0002] Due to its unique characteristics, multi-instance learning has become the fourth type of machine learning framework alongside supervised learning, unsupervised learning, and reinforcement learning, and has been successfully applied to image classification and labeling, text classification, target tracking, medical image-aided recognition, Computer security, Web page retrieval, face recognition and other fields. [0003] At present, multi-instance learning is mainly divided into multi-instance single-label learning and multi-instance multi-label learning. [0004] In multi-instance single-label learning, samples are represented as bags consisting of multiple examples, and each bag is labeled as positive or negative. A bag is positive if at leas...

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): G06K9/62
CPCG06F18/214
Inventor 邹海林陈彤彤柳婵娟丁昕苗
Owner LUDONG UNIVERSITY
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products