Research Article
Discriminative Semi-supervised Learning in Manifold Subspace for Face Recognition
@INPROCEEDINGS{10.1007/978-3-319-29236-6_24, author={Tue-Minh Vo and Hung Truong and Thai Le}, title={Discriminative Semi-supervised Learning in Manifold Subspace for Face Recognition}, proceedings={Context-Aware Systems and Applications. 4th International Conference, ICCASA 2015, Vung Tau, Vietnam, November 26-27, 2015, Revised Selected Papers}, proceedings_a={ICCASA}, year={2016}, month={4}, keywords={Face recognition Manifold Semi-supervised}, doi={10.1007/978-3-319-29236-6_24} }- Tue-Minh Vo
Hung Truong
Thai Le
Year: 2016
Discriminative Semi-supervised Learning in Manifold Subspace for Face Recognition
ICCASA
Springer
DOI: 10.1007/978-3-319-29236-6_24
Abstract
Linear Discriminant Analysis (LDA) is a commonly used method for dimensionality reduction, which preserves class separability. Despite its successes, it has limitations under some situations, including the small sample size problem. In practice, when the training data set is small, the covariance matrix of each class may not be accurately estimated. Moreover, LDA doesn’t handle unlabeled data. In this paper, we propose a semi-supervised method called Discriminative Semi-supervised Learning in Manifold subspace (DSLM), which aims at overcoming all these limitations. The proposed method is designed to explore the discriminative information of labeled data and to preserve the intrinsic geometric structure of the data. We empirically compare our method with several related methods on face databases. Results are obtained from the experiments showing the effectiveness of our proposed method .
