Spectral Clustering and Kernel PCA are Learning Eigenfunctions

In this paper, we show a direct equivalence between spectral clustering and kernel PCA, and how both are special cases of a more general learning problem, that of learning the principal eigenfunctions of a kernel, when the functions are from a Hilbert space whose inner product is defined with respect to a density model. This defines a natural mapping for new data points, for methods that only provided an embedding, such as spectral clustering and Laplacian eigenmaps. The analysis also suggests new approaches to unsupervised learning in which abstractions such as manifolds and clusters that represent the main features of the data density are extracted. Dans cet article, on montre une equivalence directe entre la classification spectrale et l'ACP a noyau, et on montre que les deux sont des cas particuliers d'un probleme plus general, celui d'apprendre les fonctions propres d'un noyau. Ces fonctions fournissent une base pour un espace de Hilbert dont le produit scalaire est defini par rapport a la densite des donnees. Les fonctions propres definissent une transformation de coordonnees naturelles pour de nouveaux points, alors que des methodes comme la classification spectrale et les 'Laplacian eigenmaps' ne fournissaient un systeme de coordonnees que pour les exemples d'apprentissage. Cette analyse suggere aussi de nouvelles approches a l'apprentissage non-supervise dans lesquelles on extrait des abstractions qui resument la densite des donnees, telles que des varietes et des classes naturelles.

[1]  R. Taylor,et al.  The Numerical Treatment of Integral Equations , 1978 .

[2]  G. Wahba Spline models for observational data , 1990 .

[3]  Shang-Hua Teng,et al.  Spectral partitioning works: planar graphs and finite element meshes , 1996, Proceedings of 37th Conference on Foundations of Computer Science.

[4]  Fan Chung,et al.  Spectral Graph Theory , 1996 .

[5]  Jitendra Malik,et al.  Normalized cuts and image segmentation , 1997, Proceedings of IEEE Computer Society Conference on Computer Vision and Pattern Recognition.

[6]  Bernhard Schölkopf,et al.  Nonlinear Component Analysis as a Kernel Eigenvalue Problem , 1998, Neural Computation.

[7]  B. Schölkopf,et al.  Advances in kernel methods: support vector learning , 1999 .

[8]  Yair Weiss,et al.  Segmentation using eigenvectors: a unifying view , 1999, Proceedings of the Seventh IEEE International Conference on Computer Vision.

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

[10]  Christopher K. I. Williams,et al.  The Effect of the Input Density Distribution on Kernel-based Classifiers , 2000, ICML.

[11]  S T Roweis,et al.  Nonlinear dimensionality reduction by locally linear embedding. , 2000, Science.

[12]  Trevor F. Cox,et al.  Metric multidimensional scaling , 2000 .

[13]  Nello Cristianini,et al.  On the Concentration of Spectral Properties , 2001, NIPS.

[14]  Mikhail Belkin,et al.  Laplacian Eigenmaps and Spectral Techniques for Embedding and Clustering , 2001, NIPS.

[15]  Michael I. Jordan,et al.  On Spectral Clustering: Analysis and an algorithm , 2001, NIPS.

[16]  Nello Cristianini,et al.  Spectral Kernel Methods for Clustering , 2001, NIPS.

[17]  P. Niyogi,et al.  Locality Preserving Projections (LPP) , 2002 .

[18]  John Shawe-Taylor,et al.  The Stability of Kernel Principal Components Analysis and its Relation to the Process Eigenspectrum , 2002, NIPS.

[19]  Pascal Vincent,et al.  Manifold Parzen Windows , 2002, NIPS.

[20]  Thomas G. Dietterich,et al.  Editors. Advances in Neural Information Processing Systems , 2002 .

[21]  Yoshua Bengio,et al.  Learning Eigenfunctions of Similarity: Linking Spectral Clustering and Kernel PCA , 2003 .

[22]  Mikhail Belkin,et al.  Laplacian Eigenmaps for Dimensionality Reduction and Data Representation , 2003, Neural Computation.

[23]  Mikhail Belkin,et al.  Semi-Supervised Learning on Riemannian Manifolds , 2004, Machine Learning.

[24]  Christopher K. I. Williams On a Connection between Kernel PCA and Metric Multidimensional Scaling , 2004, Machine Learning.