Application Of Adaptive Principal Component Extraction To Gene Expression Data

Principal Component Analysis (PCA) has been widely studied by researchers at home and broad,in which principal components are the directions in which the input data have the largest variances. As a powerful statistical technique, PCA has found wide application in many modern information processing fields, such as high resolution spectrum estimation, system identification, data compression, feature extraction, pattern recognition, digital communication, computer vision, and so on. However, the traditional PCA method requires the computation of covariance matrix and feature vectors of a stochastic vector process, which brings complexity for programming. In this dissertation we investigate the combination of artificial neural network and PCA. The emphasis is on Adaptive Principle Component Extraction (APEX) algorithm, which is able to extract in parallel multiple eigen-components of a covariance matrix.In this paper, APEX algorithm and traditional PCA algorithm were applied to artificial data in order to compare their performance of extracting eigen vectors and eigen values.Then, APEX algorithm is applied to real gene expression data, which has rarely involved in other literatures. Experiments show that the parallelizable APEX algorithm significantly outperforms traditional PCA method in terms of convergence speed. Data reconstruction is then made according to the primary features, which lays the pavement for further research.