Abstract An adaptive neighborhood selection algorithm is proposed based on deflection angle of local tangent space by using the geometric properties of local tangent space. It computes the angle between local centralized samples and its tangent space based on the orthogonal projection of local tangent space. It depicts the properties of local tangent space better, and discriminates the samples which do not belong to this neighborhood and possesses better antinoise ability. The proposed algorithm is a modification to local tangent space alignment with manifold learning function of local high curvature. Experimental results show that the proposed algorithm is effective.
[1] Jolliffe I T. Principal Component Analysis. New York, USA: Springer-Verlag, 1986 [2] Zhang Zhenyue, Zha Hongyuan. Principal Manifolds and Nonlinear Dimensionality Reduction via Tangent Space Alignment. SIAM Journal of Scientific Computing, 2005, 26(1): 313-338 [3] Tenenbaum J B, de Silva V, Langford J C. A Global Geometric Framework for Non-Linear Dimension Reduction. Science, 2000, 290(5500): 2319-2323 [4] Cox T, Cox M. Multidimensional Scaling. London, UK: Chapman Hall, 1994 [5] Wang Jing, Zhang Zhenyue, Zha Hongyuan. Adaptive Manifold Learning // Saul L K, Weiss Y, Bottou L, eds. Neural Information Processing Systems. Cambridge, USA: MIT Press, 2004, XVII: 1473-1480 [6] Min Wanli, Lu Ke, He Xiaofei. Locality Pursuit Embedding. Pattern Recognition, 2004, 37(4): 781-788 [7] Roth P M, Winter M. Survey of Appearance-Based Methods for Object Recognition. Technical Report, ICA-TR-01/08, Graz, Austria: University of Technology. Institute for Computer Graphics and Vision, 2008 [8] Golub G H, van Loan C F. Matrix Computations. 3rd Edition. London, UK: Johns Hopkins University Press, 1996
2010, Vol. 23 
