Semi-supervised feature selection for gender classification

摘要: We apply a semi-supervised learning method to perform gender determination. The aim is select the most discriminating feature components from eigen-feature representation of faces. By making use information provided by both labeled and unlabeled data, we successfully reduce size data set required for selection, improve classification accuracy. Instead using 2D brightness images, 2.5D facial needle-maps which reveal more directly shape information. Principal geodesic analysis (PGA), generalization principal component (PCA) residing in Euclidean space on manifold, used obtain needle-maps. In our experiments, achieve 90.50% accuracy when 50% are labeled. This performance demonstrates effectiveness this small set, feasibility