Feature Selection for Identifying Critical Variables of Principal Components Based on K-Nearest Neighbor Rule

摘要

Principal components analysis (PCA) is a popular linear feature extractor to unsupervised dimensionality reduction, and found in many branches of science including-examples in computer vision, text processing and bioinformatics, etc. However, axes of the lower-dimensional space, i.e., principal components, are a set of new variables carrying no clear physical meanings. Thus, interpretation of results obtained in the lower-dimensional PCA space and data acquisition for test samples still involve all of the original measurements. To select original features for identifying critical variables of principle components, we develop a new method with k-nearest neighbor clustering procedure and three new similarity measures to link the physically meaningless principal components back to a subset of original measurements. Experiments are conducted on benchmark data sets and face data sets with different poses, expressions, backgrounds and occlusions for gender classification to show their superiorities.