Input Space Partitioning for Neural Network Learning

Input Space Partitioning for Neural Network Learning

Shujuan Guo (School of Electronic & Information Engineering, Xi’an Jiaotong University, Suzhou, Jiangsu, China), Sheng-Uei Guan (School of Electronic & Information Engineering, Xi’an Jiaotong University, Suzhou, Jiangsu, China), Weifan Li (Dept. of Computer Science and Software Engineering, Xi’an Jiaotong-Liverpool University, Suzhou, China), Ka Lok Man (Dept. of Computer Science and Software Engineering, Xi’an Jiaotong-Liverpool University, Suzhou, China), Fei Liu (Department of Computer Science & Computer Engineering, La Trobe University, Melbourne, VIC, Australia) and A. K. Qin (School of Computer Science and Information Technology, RMIT University, Melbourne, VIC, Australia)
Copyright: © 2013 |Pages: 11
DOI: 10.4018/jaec.2013040105
OnDemand PDF Download:
$30.00
List Price: $37.50

Abstract

To improve the learning performance of neural network (NN), this paper introduces an input attribute grouping based NN ensemble method. All of the input attributes are partitioned into exclusive groups according to the degree of inter-attribute promotion or correlation that quantifies the supportive interactions between attributes. After partitioning, multiple NNs are trained by taking each group of attributes as their respective inputs. The final classification result is obtained by integrating the results from each NN. Experimental results on several UCI datasets demonstrate the effectiveness of the proposed method.
Article Preview

Terminology And Concepts

Let denote the promotion rate of two attributes i and j, which is defined by:

(1) where represents the classification error obtained by training with single attribute i, and represents the classification error obtained by training with two attributes i and j. When the promotion rate of two attributes is 1, these two attributes are considered to be mutually supportive for classification. Otherwise, they are considered to be mutually interfered for classification.

To take full advantage of inter-attribute promotions, we compute the average value of all with Any pair of attributes whose classification error is less than this average value is considered to have significant promotion to each other. The smaller the corresponding classification error is, the more significant the promotion is.

In statistics, correlation measures the strength and direction of the linear relationship between two random variables. There exist many ways of calculating correlation. This paper employs the Pearson’s correlation coefficients (Sedgwick, 2012).

Complete Article List

Search this Journal:
Reset
Open Access Articles: Forthcoming
Volume 7: 4 Issues (2016): 2 Released, 2 Forthcoming
Volume 6: 4 Issues (2015)
Volume 5: 4 Issues (2014)
Volume 4: 4 Issues (2013)
Volume 3: 4 Issues (2012)
Volume 2: 4 Issues (2011)
Volume 1: 4 Issues (2010)
View Complete Journal Contents Listing