Local relative transformation with application to isometric embedding

Current isometric embedding approaches are topologically unstable when confronted with noisy data, as where the neighborhood is critically distorted. Based on the cognitive law, a relative transformation (RT), which improves the distinction between data points and diminishes the impact of noise on isometric embedding approaches, is proposed. As the constructed space from large scale data by RT is high-dimensional, local relative transformation (LRT) is further proposed. Subsequently, a new isometric embedding approach is developed by using LRT to construct a better neighborhood graph with fewer short-circuit edges, while the embedding is still performed in the original space. This approach has significantly increased performance and reduced running time. The proposed approach was validated by experiments on challenging benchmark data sets.

[1]  Anil K. Jain,et al.  Incremental nonlinear dimensionality reduction by manifold learning , 2006, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[2]  Li Yang.  Building k edge-disjoint spanning trees of minimum total length for isometric data embedding , 2005, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[3]  Lijun Jiang,et al.  Globalizing Local Neighborhood for Locally Linear Embedding , 2006, 2006 IEEE International Conference on Systems, Man and Cybernetics.

[4]  Paul L. Rosin,et al.  Selection of the optimal parameter value for the Isomap algorithm , 2006, Pattern Recognit. Lett..

[5]  Robert Pless,et al.  Image distance functions for manifold learning , 2007, Image Vis. Comput..

[6]  Lijun Jiang,et al.  Performing Locally Linear Embedding with Adaptable Neighborhood Size on Manifold , 2006, PRICAI.

[7]  Lijun Jiang,et al.  Clustering-based Locally Linear Embedding , 2006, 2006 IEEE International Conference on Systems, Man and Cybernetics.

[8]  Joshua B. Tenenbaum,et al.  The Isomap Algorithm and Topological Stability , 2002, Science.

[9]  Amitabha Mukerjee,et al.  Non-linear Dimensionality Reduction by Locally Linear Isomaps , 2004, ICONIP.

[10]  Tim W. Nattkemper,et al.  ISOLLE: LLE with geodesic distance , 2006, Neurocomputing.

[11]  Joshua B. Tenenbaum,et al.  Global Versus Local Methods in Nonlinear Dimensionality Reduction , 2002, NIPS.

[12]  J. Tenenbaum,et al.  A global geometric framework for nonlinear dimensionality reduction. , 2000, Science.

[13]  Li Yang.  Building k-connected neighborhood graphs for isometric data embedding , 2006, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[14]  D. Donoho,et al.  Hessian eigenmaps: Locally linear embedding techniques for high-dimensional data , 2003, Proceedings of the National Academy of Sciences of the United States of America.

[15]  Zhi-Hua Zhou,et al.  Supervised nonlinear dimensionality reduction for visualization and classification , 2005, IEEE Transactions on Systems, Man, and Cybernetics, Part B (Cybernetics).

[16]  Ann B. Lee,et al.  Diffusion maps and coarse-graining: a unified framework for dimensionality reduction, graph partitioning, and data set parameterization , 2006, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[17]  Lijun Jiang,et al.  Clustering-Based Nonlinear Dimensionality Reduction on Manifold , 2006, PRICAI.

[18]  Lijun Jiang,et al.  Using Graph Algebra to Optimize Neighborhood for Isometric Mapping , 2007, IJCAI.

[19]  Heeyoul Choi,et al.  Robust kernel Isomap , 2007, Pattern Recognit..