Manifold learning is proved to be an efficient tool for nonlinear dimensionality reduction. Various local and global learners have been proposed to successfully extract intrinsic geometry underlying high-dimensional data cloud. However, there is no work considering the ensemble of local and global manifold learners to promote learning results, where such strategy has achieved great success in classification. In this paper, we propose a manifold learner ensemble method (MLEN) for the first time. MLEN consists of a local manifold learner and a global one. By fusing the extracted local and global geometry, MLEN outperforms each of its components and outputs an overall and superior embedding. Experimental results on both synthetic and image manifolds validate the effectiveness of the proposed method. © Springer-Verlag Berlin Heidelberg 2013.
CITATION STYLE
Zhang, P., Fan, C., Ren, Y., & Zhang, N. (2013). Manifold learner ensemble. In Communications in Computer and Information Science (Vol. 375, pp. 94–99). Springer Verlag. https://doi.org/10.1007/978-3-642-39678-6_16
Mendeley helps you to discover research relevant for your work.