A Supervised Feature Extraction Algorithm for Multi-class

  • Authors:

  • Shifei Ding;Fengxiang Jin;Xiaofeng Lei;Zhongzhi Shi

  • Affiliations:

  • School of Computer Science and Technology, China University of Mining and Technology, Xuzhou, China 221008 and Key Laboratory of Intelligent Information Processing, Institute of Computing Technolo ...;College of Geoinformation Science and Engineering, Shandong University of Science and Technology, Qingdao, P.R. China 266510;School of Computer Science and Technology, China University of Mining and Technology, Xuzhou, China 221008;Key Laboratory of Intelligent Information Processing, Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China 100080

  • Venue:
  • FAW '08 Proceedings of the 2nd annual international workshop on Frontiers in Algorithmics
  • Year:
  • 2008

Quantified Score

Hi-index 0.00

Visualization

Abstract

In this paper, a novel supervised information feature extraction algorithm is set up. Firstly, according to the information theories, we carried out analysis for the concept and its properties of the cross entropy, then put forward a kind of lately concept of symmetry cross entropy (SCE), and point out that the SCE is a kind of distance measure, which can be used to measure the difference of two random variables. Secondly, Based on the SCE, the average symmetry cross entropy (ASCE) is set up, and it can be used to measure the difference degree of a multi-class problem. Regarding the ASCE separability criterion of the multi-class for information feature extraction, a novel algorithm for information feature extraction is constructed. At last, the experimental results demonstrate that the algorithm here is valid and reliable, and provides a new research approach for feature extraction, data mining and pattern recognition.