Proceedings of the 13th EAI International Conference on Mobile Multimedia Communications, Mobimedia 2020, 27-28 August 2020, Cyberspace

Research Article

Discriminative Graph Learning via Low-Rank Constraint for Semi-Supervised Learning

Download389 downloads
  • @INPROCEEDINGS{10.4108/eai.27-8-2020.2296560,
        author={Wang  Ying and Ao  Li},
        title={Discriminative Graph Learning via Low-Rank Constraint for Semi-Supervised Learning},
        proceedings={Proceedings of the 13th EAI International Conference on Mobile Multimedia Communications, Mobimedia 2020, 27-28 August 2020, Cyberspace},
        publisher={EAI},
        proceedings_a={MOBIMEDIA},
        year={2020},
        month={11},
        keywords={image classification graph learning low-rank constraints semi-supervised},
        doi={10.4108/eai.27-8-2020.2296560}
    }
    
  • Wang Ying
    Ao Li
    Year: 2020
    Discriminative Graph Learning via Low-Rank Constraint for Semi-Supervised Learning
    MOBIMEDIA
    EAI
    DOI: 10.4108/eai.27-8-2020.2296560
Wang Ying1, Ao Li1,*
  • 1: School of Computer Science and Technology, Harbin University of Science and Technology, Harbin, China
*Contact email: dargonboy@126.com

Abstract

Graph learning is more and more widely used in discovering associations between data and mining relationships between data. Our paper proposed a discriminative graph learning for semi-supervised learning method discriminative graph learning with low rank constraints. Traditional semi-supervised learning methods exist the problem of similarity between similar species is not high. How to improve the similarity between similar species has become a problem we need to solve in this field. In order to solve this problem, this paper proposed a new semi-supervised learning means, which mainly includes two parts: 1) Representing the local structure of the research sample by low-rank representation, compressing similar content, and easily finding the focus of image classification; 2) in order to improve the similarity between classes, A term that uses similarity constraints between the data to be classified and a known model. In our paper, the proposed means is used for comparative experiments of classification on multiple data sets. Through experiments show that the means proposed in this paper performs better than traditional semi-supervised classification methods in terms of accuracy of classification.