A Semi-supervised Dimensionality Reduction Framework for Face Recognition Based on Sparse Lorentzian Metric Tensors

摘要

There has been significant recent interest in extending supervised algorithms to semi-supervised form which preserve local structures of the unlabeled samples. However, how to choose the homogeneous points is still an open-problem. In this paper, by introducing the sparse Lorentzian metric, we propose a general framework to extend supervised algorithms to semi-supervised case. Our proposed techniques can find the homogeneous points of the unlabeled samples in a more natural way. The learnt sparse Lorentzian metric tensors can also keep both the local structure of the unlabeled samples and their global geometrical structure. The experimental results on face recognition show that our algorithm achieves better recognition accuracy.