Dimension reduction and detection of outliers in cancer classification using information complexity for undersized samples

Recent developments in DNA microarray techniques has allowed simultaneously to display thousands of the potential gene expressions. Due to the wealth of gene expression data, researchers have begun to focus their attention on how to optimally classify cancer using the gene expression data. Although many of the methods used have produced promising results, there remains many problems yet to be resolved and to be understood. One of the most important of these problems is the dimension reduction and the detection of outliers. The classical statistical techniques cannot be used to reduce the dimension and to detect the outliers because of the severity of undersized sample problem (n<