Transfer learning method based on paired sample matching

A sample matching and transfer learning technology, applied in the field of image classification and transfer learning, can solve problems such as the decline of the effect, and achieve the effect of enhancing the difference, improving the generalization ability, and fully training

Active Publication Date: 2019-12-10
SHANDONG COMP SCI CENTNAT SUPERCOMP CENT IN JINAN
View PDF13 Cites 11 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0007] The purpose of the present invention is to solve the problem that the effect of the commonly used methods decreases when there are fewer target samples in the transfer learning based on double chains, and to provide an efficient transfer learning method based on paired sample matching, which is used to make full use of the source domain and Association of samples from the target domain and validated in image classification 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
  • Transfer learning method based on paired sample matching
  • Transfer learning method based on paired sample matching
  • Transfer learning method based on paired sample matching

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0056] Such as figure 1 As shown, the operation steps of the method include:

[0057] Step 10 Data Preprocessing

[0058] Such as figure 2 and image 3 As shown, the mixed National Institute of Standards and Technology dataset (MNIST) and the United States Postal Service dataset (USPS) are commonly used datasets for transfer learning. They contain images of numbers from 0 to 9. Two cross-domain tasks, MNIST→USPS and USPS→MNIST, are considered, and 2000 images in MNIST and 1800 images in USPS are randomly selected. Treat each image as a sample, pair it with other samples, and split into positive and negative pairs. When the number of samples per class in the target domain is n, for the MNIST→USPS task, there are 2000*n positive samples and 18000*n negative samples; for the USPS→MNIST task, there are 1800*n positive samples and 16200*n negative samples. Each task was repeated 10 times to obtain the average value.

[0059] Such as Figure 4 As shown, the Office-31 datas...

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 belongs to the technical field of image classification and transfer learning, discloses a transfer learning method based on paired sample matching, and realizes mining of internal relations of samples based on different domains. The method specifically comprises the following steps of (1) data preprocessing, (2) double-chain model construction based on transfer learning, (3) instancenormalization and batch normalization, and (4) calculation of comparison loss and maximum mean distance loss. The method has the advantages that instance normalization and batch normalization are combined for learning at the same time, styles and semantic association characteristics of different images are fully mined, and efficient recognition of a small number of target domain samples under theassistance of a source domain is achieved.

Description

technical field [0001] The invention belongs to the technical field of image classification and transfer learning, and relates to a transfer learning method based on paired sample matching, which is used to mine the inherent relevance of different samples. In the field of image classification, the method of paired sample matching and transfer learning is verified. effectiveness. Background technique [0002] Deep convolutional neural networks are widely used in various machine learning scenarios, such as image recognition, object detection, and semantic segmentation. Unfortunately, many existing methods are usually only applicable to specific domains and rely on data with a large number of labels. If the data in the target domain is unavailable or difficult to label, the effect of traditional machine learning methods will drop significantly. To address this problem, a common approach is to use transfer learning and domain adaptation to learn a discriminative and domain-inva...

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): G06F16/2458G06K9/62G06N3/04
CPCG06F16/2465G06N3/044G06N3/045G06F18/214
Inventor 高赞李荫民程志勇陈达舒明雷聂礼强
Owner SHANDONG COMP SCI CENTNAT SUPERCOMP CENT IN JINAN
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