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

An unsupervised cross-domain adaptive data calibration method and system based on weighted distribution alignment and geometric feature alignment

A technology of geometric features and calibration methods, applied in the field of data calibration, which can solve the problems of not explicitly considering the distribution differences between fields, the distortion or stretching of the original feature space, and the lack of further utilization of geometric features.

Inactive Publication Date: 2019-04-16
HARBIN INST OF TECH
View PDF0 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

The domain adaptation method based on feature transformation can be divided into data-centered method and subspace-centered method. The main purpose of the data-centered method is to find a consistent transformation to map the data of the source domain and the target domain. To a domain invariant space to reduce the distribution difference and maintain the data characteristics of the original space, but this method does not further utilize the geometric characteristics of the data, because the original feature space has been distorted or stretched after the feature transformation; the subspace is used as The central method only deals with subspaces, and does not explicitly consider the distribution differences between domains after mapping

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
  • An unsupervised cross-domain adaptive data calibration method and system based on weighted distribution alignment and geometric feature alignment
  • An unsupervised cross-domain adaptive data calibration method and system based on weighted distribution alignment and geometric feature alignment
  • An unsupervised cross-domain adaptive data calibration method and system based on weighted distribution alignment and geometric feature alignment

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0097] The unsupervised cross-domain adaptive data calibration method and system based on weighted distribution alignment and geometric feature alignment described in the present invention are described as follows in conjunction with the drawings and tables:

[0098] As the name implies, weighted distribution alignment and geometric feature alignment are to weight and align the marginal probability distribution and conditional probability distribution of the data and to align the geometric features of the sample data spatial features, while introducing Laplacian regularization terms to further maintain the sample data. The geometric structure of the space finally obtains an optimization problem that can get a closed-form solution, thus solving the domain adaptation problem. The system is tested on a large number of public data sets of image recognition and character recognition, and the proposed method based on Effectiveness of Cross-Domain Unsupervised Domain Adaptive Data Cal...

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 an unsupervised cross-domain adaptive data calibration method and system based on weighted distribution alignment and geometric feature alignment, and relates to the technicalfield of data calibration. The invention aims to effectively improve the data calibration accuracy. Weighted distribution alignment can balance the importance of marginal probability distribution andconditional probability distribution of the sample data, so that the difference between domains is reduced. The geometric feature alignment not only can further mine the geometric features of the sample data between the fields, but also can well maintain the geometric structure of the sample data space through the Grabbs regularization, thereby improving the sample separability and the data calibration accuracy. Compared with other methods for experimental comparison, the system-unsupervised cross-domain adaptive data calibration method based on weighted distribution alignment and geometric feature alignment can effectively improve the data calibration accuracy.

Description

technical field [0001] The invention relates to an unsupervised cross-domain adaptive data calibration method and system, and relates to the technical field of data calibration. Background technique [0002] The unsupervised domain adaptation problem is a sub-problem of transfer learning, which aims to solve the domain adaptation problem without labeled data in the target domain. Previous research results mainly focus on sample-based domain adaptation and feature transformation-based domain adaptation. The domain adaptation method based on feature transformation can be divided into data-centered method and subspace-centered method. The main purpose of the data-centered method is to find a consistent transformation to map the data of the source domain and the target domain. To a domain invariant space to reduce the distribution difference and maintain the data characteristics of the original space, but this method does not further utilize the geometric characteristics of the...

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
IPC IPC(8): G06N7/00
CPCG06N7/00
Inventor 何慧张伟哲方滨兴杨洪伟李韬白雅雯
Owner HARBIN INST OF TECH
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

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

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com