Graph embedding based feature selection

Usually many real datasets in pattern recognition applications contain a large quantity of noisy and redundant features that are irrelevant to the intrinsic characteristics of the dataset. The irrelevant features may seriously deteriorate the learning performance. Hence feature selection which aims to select the most informative features and to eliminate the irrelevant feature from the original dataset plays an important role in data mining, image recognition and microarray data analysis. In this paper, we developed a new feature selection technique based on the recently developed graph embedding framework for manifold learning. We propose a recursive feature elimination (RFE) method using feature score for identifying the optimal feature subset. One advantage of the RFE method is that it can successfully identify the nonlinear features based on manifold learning. The experimental results both on face dataset and microarray dataset verify the effectiveness and efficiency of the proposed method.