Informatics and Applications
2018, Volume 12, Issue 1, pp 71-77
PRINCIPAL AXES RECONSTRUCTION
Abstract
Principal component analysis (PCA) is a widely used technique for processing, compressing, and
visualizing of data. New possibilities are opened by probabilistic PCA (PPCA), realized within the maximum
likelihood principle for a Gaussian model with latent variables. Within the framework of PPCA, data processing
algorithms have appeared, aimed at reducing the dimensionality of data and providing the transition to the space of
the main components, but not explicitly giving the characteristics of the main components. The article is devoted to
details that deepen the understanding of the features of PPCA and corrections of the errors revealed in publications.
Two methods for reconstructing the characteristics of principal components are proposed and substantiated. One
of them is based on recalculation of the covariance matrix in the formed space of main components. The other
method consists in successively repeating the same steps: identifying the first main component and excluding it
from data analysis.
[+] References (8)
- Jolliffe, I. T 2002. Principal component analysis. New York,
NY: Springer. 487 p.
- Gabriel, K. R. 1978. Least squares approximation of matrices by additive and multiplicative models. J. Roy. Stat.
Soc. B 40(2):186-196.
- Tipping, M. E., and C. M. Bishop. 1999. Mixtures of probabilistic principal component analyzers. Neural Comput.
11(2):443-482.
- Rubin, D.B., and D. T Thayer. 1982. EM algorithms for
ML factor analysis. Psychometrika 47(1):69-76.
- Roweis, S. 1997. EM algorithm for PCA and SPCA. Advances in neural information processing systems. Eds. M.I. Jordan, M. J. Kearns, and S. A. Solla. MIT Press. 10:626-632.
- Ahn, J. H., and J. H. Oh. 003. A constrained EM algorithm for principal component. Neural Comput. 15:57-65.
- Wang, H., and Z. Hu. 2007. An unified EM algorithm for PCA and KPCA. Neurocomputing 71:459-462.
- Choi, S., J. H. Ahn, and A. Cichocki. 2006. Constrained projection approximation algorithms for principal component analysis. Neural Process. Lett.
[+] About this article
Title
PRINCIPAL AXES RECONSTRUCTION
Journal
Informatics and Applications
2018, Volume 12, Issue 1, pp 71-77
Cover Date
2018-03-30
DOI
10.14357/19922264180109
Print ISSN
1992-2264
Publisher
Institute of Informatics Problems, Russian Academy of Sciences
Additional Links
Key words
principal component analysis; EM-algorithm; reconstruction of axes and dispersion
Authors
M. P. Krivenko 
Author Affiliations
Institute of Informatics Problems, Federal Research Center "Computer Science and Control" of the Russian Academy of Sciences, 44-2 Vavilov Str., Moscow 119333, Russian Federation
|