Supervised feature extraction algorithm based on continuous divergence criterion

  • Authors:

  • Shifei Ding;Zhongzhi Shi;Fengxiang Jin

  • Affiliations:

  • College of Information Science and Eng., Shandong Agricultural Univ., Taian, P.R. China and Key Laboratory of Intelligent Information Processing, Institute of Computing Technology, Chinese Academy ...;Key Laboratory of Intelligent Information Processing, Institute of Computing Technology, Chinese Academy of Sciences, Beijing, P.R. China;College of Geo-Information Science and Engineering, Shandong University of Science and Technology, Qingdao, P.R. China

  • Venue:
  • ICIC'06 Proceedings of the 2006 international conference on Intelligent computing: Part II
  • Year:
  • 2006

Quantified Score

Hi-index 0.00

Visualization

Abstract

Feature extraction plays an important part in pattern recognition (PR), data mining, machine learning et al. In this paper, a novel supervised feature extraction algorithm based on continuous divergence criterion (CDC) is set up. Firstly, the concept of the CDC is given, and some properties of the CDC are studied, and proved that CDC here is a kind of distance measure, i.e. it satisfies three requests of distance axiomatization, which can be used to measure the difference degree of a two-class problem. Secondly, based on CDC, the basic principle of supervised feature extraction are studied, a new concept of accumulated information rate (AIR) is given, which can be used to measure the degree of feature compression for two-class, and a new supervised feature extraction algorithm is constructed. At last, the experimental results demonstrate that the algorithm here is valid and reliable.