Orthogonal locality discriminant embedding for document classification
Ziqiang Wang, Xia Sun · 2009
Dimensionality reduction algorithms, which try to reduce the dimensionality of data and to enhance the discriminant information, are of paramount importance in document classification. In this paper, a novel dimensionality reduction algorithm called orthogonal locality discriminant embedding (OLDE) is proposed to address these problems. The OLDE algorithm effectively combines the idea of local discriminant embedding (LDE) and orthogonal basis functions, which utilizes both the local manifold structure and label information to enhance discriminative power. Extensive experiments on three document databases show the effectiveness of the proposed OLDE algorithm.