Abstract
We present an extension of Principal Component Analysis (PCA) and a new algorithm for clustering points in Rn based on it. The key property of the algorithm is that it is affine-invariant. When the input is a sample from a mixture of two arbitrary Gaussians, the algorithm correctly classifies the sample assuming only that the two components are separable by a hyperplane, i.e., there exists a halfspace that contains most of one Gaussian and almost none of the other in probability mass. This is nearly the best possible, improving known results substantially [15, 10, 1]. For k>2 components, the algorithm requires only that there be some (k−1)-dimensional subspace in which the overlap in every direction is small. Here we define overlap to be the ratio of the following two quantities: 1) the average squared distance between a point and the mean of its component, and 2) the average squared distance between a point and the mean of the mixture. The main result may also be stated in the language of linear discriminant analysis: if the standard Fisher discriminant [9] is small enough, labels are not needed to estimate the optimal subspace for projection. Our main tools are isotropic transformation, spectral projection and a simple reweighting technique. We call this combination isotropic PCA.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
References
D. Achlioptas and F. McSherry, On spectral learning of mixtures of distributions, in: Proc. of COLT, 2005.
K. Chaudhuri and S. Rao, Beyond gaussians: Spectral methods for learning mixtures of heavy-tailed distributions (2008).
K. Chaudhuri and S. Rao, Learning mixtures of product distributions using corre-lations and independence (2008).
A. Dasgupta, J. Hopcroft, J. Kleinberg and M. Sandler, On learning mixtures of heavy-tailed distributions, in: Proc. of FOCS (2005).
S. DasGupta, Learning mixtures of gaussians, in: Proc. of FOCS (1999).
A. P. Dempster, N. M. Laird and D. B. Rubin, Maximum likelihood from incom-plete data via the em algorithm, Journal of the Royal Statistical Society B, 39 (1977), 1–38.
Jon Feldman, Rocco A. Servedio and Ryan O’Donnell, Pac learning axis-aligned mixtures of gaussians with no separation assumption, in: COLT (2006), pp. 20–34.
K. Fukunaga, Introduction to Statistical Pattern Recognition, Academic Press (1990).
R. O. Duda, P. E. Hart and D. G. Stork, Pattern Classification, John Wiley & Sons (2001).
R. Kannan, H. Salmasian and S. Vempala, The spectral method for general mixture models, in: Proceedings of the 18th Conference on Learning Theory, University of California Press (2005).
L. Lovász and S. Vempala, The geometry of logconcave functions and and sampling algorithms, Random Strucures and Algorithms, 30(3) (2007), 307–358.
J. B. MacQueen, Some methods for classification and analysis of multivariate ob-servations, in: Proceedings of 5-th Berkeley Symposium on Mathematical Statistics and Probability, volume 1. University of California Press (1967), pp. 281–297.
M. Rudelson, Random vectors in the isotropic position, Journal of Functional Analysis, 164 (1999), 60–72.
M. Rudelson and R. Vershynin, Sampling from large matrices: An approach through geometric functional analysis, J. ACM, 54(4) (2007).
R. Kannan and S. Arora, Learning mixtures of arbitrary gaussians, Ann. Appl. Probab., 15(1A) (2005), 69–92.
L. Schulman and S. DasGupta, A two-round variant of em for gaussian mixtures, in: Sixteenth Conference on Uncertainty in Artificial Intelligence (2000).
G. W. Stewart and Ji guang Sun, Matrix Perturbation Theory, Academic Press, Inc. (1990).
S. Vempala and G. Wang, A spectral algorithm for learning mixtures of distributions, Proc. of FOCS 2002; JCCS, 68(4) (2004), 841–860.
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2008 János Bolyai Mathematical Society and Springer-Verlag
About this chapter
Cite this chapter
Brubaker, S., Vempala, S.S. (2008). Isotropic PCA and Affine-Invariant Clustering. In: Grötschel, M., Katona, G.O.H., Sági, G. (eds) Building Bridges. Bolyai Society Mathematical Studies, vol 19. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-540-85221-6_8
Download citation
DOI: https://doi.org/10.1007/978-3-540-85221-6_8
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-85218-6
Online ISBN: 978-3-540-85221-6
eBook Packages: Mathematics and StatisticsMathematics and Statistics (R0)