This research topic is described in Section 2.6 of the Bayes chapter of the biennial report 2008-2009, available here.
We have applied the rectified factor analysis method (Harva and Kabán, 2007) to the analysis of real stellar population spectra of elliptical galaxies in (Nolan et al., 2006). Ellipticals are the oldest galactic systems in the local universe and have been studied well in astrophysics. The hypothesis that some of these old galactic systems may actually contain young components is relatively new. Hence, we have investigated whether a set of stellar population spectra can be decomposed and explained in terms of a small set of unobserved spectral prototypes in a data driven but yet physically meaningful manner. The positivity constraint is important in this modelling application, as negative values of flux would not be physically interpretable. Using a set of 21 real stellar population spectra, we found that they can indeed be decomposed to prototypical spectra, especially to a young and old component (Nolan et al., 2006).
The left figure above shows a spectrum decomposed by our algorithm to an old and an young component. The right figure illustrates the correlation between the weight of the dominant spectrum and the age of the galaxy.
M. Harva, A. Kabán. "Variational Learning for Rectified
Factor Analysis". Signal Processing, vol. 87, no. 3, 2007,
pp. 509-527.
Pdf (655k)
L. Nolan, M. Harva, A. Kaban, and S. Raychaudhury, "A data-driven Bayesian approach for finding young stellar populations in early-type galaxies from their UV-optical spectra". Monthly Notices of the Royal Astronomical Society, vol. 366, no. 1, 2006, pp. 321-338. Gzipped postscript (279k).