ISABoost: A weak classifier inner structure adjusting based AdaBoost algorithm-ISABoost based application in scene categorization

  • Authors:

  • Xueming Qian;Yuan Yan Tang;Zhe Yan;Kaiyu Hang

  • Affiliations:

  • School of Electronic and Information Engineering, Xi'an Jiaotong University, Xianning Road, Xi'an, China and Faculty of Science and Technology, The University of Macau, Macau, China;Faculty of Science and Technology, The University of Macau, Macau, China;School of Electronic and Information Engineering, Xi'an Jiaotong University, Xianning Road, Xi'an, China;School of Electronic and Information Engineering, Xi'an Jiaotong University, Xianning Road, Xi'an, China

  • Venue:
  • Neurocomputing
  • Year:
  • 2013

Quantified Score

Hi-index 0.01

Visualization

Abstract

AdaBoost algorithms fuse weak classifiers to be a strong classifier by adaptively determine fusion weights of weak classifiers. In this paper, an enhanced AdaBoost algorithm by adjusting inner structure of weak classifiers (ISABoost) is proposed. In the traditional AdaBoost algorithms, the weak classifiers are not changed once they are trained. In ISABoost, the inner structures of weak classifiers are adjusted before their fusion weights determination. ISABoost inherits the advantages of the AdaBoost algorithms in fusing weak classifiers to be a strong classifier. ISABoost gives each weak classifier a second chance to be adjusted stronger. The adjusted weak classifiers are more contributive to make correct classifications for the hardest samples. To show the effectiveness of the proposed ISABoost algorithm, its applications in scene categorization are evaluated. Comparisons of ISABoost and AdaBoost algorithms on three widely utilized scene datasets show the effectiveness of ISABoost algorithm.