Robust Nonlinear Dimension Reduction: A Self-organizing Approach
Most NDR algorithms need to solve large-scale eigenvalue problems or some variation of eigenvalue problems, which is of quadratic complexity of time and might be unpractical in case of large-size data sets. Besides, current algorithms are global, which are often sensitive to noise and disturbed by ill-conditioned matrix. In this paper, we propose a novel self-organizing NDR algorithm: SIE. The time complexity of SIE is O(NlogN). The main computing procedure of SIE is local, which improves the robustness of the algorithm remarkably.
KeywordsAnchor Point Geodesic Distance Noise Point Outer Cycle Swiss Data
Unable to display preview. Download preview PDF.
- 1.Baum, E.: What Is Thought? MIT Press, Cambridge (2004)Google Scholar
- 2.Tenenbaum, J.B., et al.: Science, December 22, vol. 290 (2000)Google Scholar
- 3.Roweis, S.T., et al.: Science, December 22, vol. 290 (2000)Google Scholar
- 4.Belkin, M., et al.: NIPS (2001)Google Scholar
- 5.Donoho, D.L., et al.: PNAS, May 13, vol. 100(10) (2003)Google Scholar
- 6.Agrafiotis, D.K., et al.: A self-organizing principle for learning nonlinear manifolds. PNAS Early Edition. Google Scholar
- 8.Yuexian, H., Zheng, D., Pilian, H.: Self-organizing Isometric Embedding. Computer research and development (2) (2005) (in Chinese)Google Scholar