Computer and Modernization

Previous Articles     Next Articles

Dictionary Learning Algorithm Based on Label Consistent and Locality Constraint

  

  1. (1. Industrial Training Center, Guangdong Polytechnic Normal University, Guangzhou 510665, China;
    2. Bio-Computing Research Center, Shenzhen Graduate School, Harbin Institute of Technology, Shenzhen 518055, China)
  • Received:2016-12-21 Online:2017-06-23 Published:2017-06-23

Abstract: In order to improve the classification performance of the learned dictionary, a dictionary learning algorithm based on the label consistent and locality constraint of atoms (LCLCDL) was proposed. The discriminative sparse codes matrix was constructed by using the labels of the atoms and training samples, and then the label consistent model was designed as the discriminative term. It can encourage the training samples of the same class to have similar coding coefficients. The locality constraint model was constructed by using the atoms and the lines of coding coefficient matrix (Profiles), and it can inherit the geometric structure of the training samples. Experiment results show that the LCLCDL algorithm achieves more higher classification performance than five sparse coding and dictionary learning algorithms.

Key words: dictionary learning, label consistent, property of locality, image classification

CLC Number: