©1996-2019 All Rights Reserved. Online Journal of Bioinformatics. You may not store these pages in any form except for your own personal use. All other usage or distribution is illegal under international copyright treaties. Permission to use any of these pages in any other way besides the before mentioned must be gained in writing from the publisher. This article is exclusively copyrighted in its entirety to OJB publications. This article may be copied once but may not be reproduced or re-transmitted without the express permission of the editors.
Online Journal of Bioinformatics©
Volume 6 (1):91-98, 2005.
Shen Li, Tan EC.
School of Computer Engineering, Nanyang Technological University, Singapore, 639798.
Li S, Tan EC., Combining Kernel Dimension-Reduction with Regularized Classifiers for Tissue Categorization, Onl J Bioinform., 6 (1):91-98, 2005. Microarrays can be used for accurate cancer diagnosis using machine learning techniques. However, the number of samples (arrays) are significantly smaller than the number of variables (genes) invalidating traditional statistical methods. The application of dimension reduction methods have achieved good performance. In this paper, two kernel dimension-reduction methods, KPLS and KPCA are considered. After dimension reduction, a linear RLSC is followed up for classification. Thus two classifiers, KPLS+RLSC and KPCA+RLSC have been derived and tested on several cancer datasets. It has been illustrated that by using mutual orthogonality of the components from kernel dimension reduction, a very simple solution can be derived for the linear RLSC, which only requires the inversion of a sparse and low-dimensional matrix. The two algorithms have also been shown to have excellent performance in comparison with SVM. Several other issues regarding the usage of KPLS and KPCA are discussed and could be useful when combining them with other classification methods.
Key words: kernel, dimension, classifiers, tissue.