Searcharxiv⌕ Search
Home/Archive/arXiv

arXiv · 0709.0928

Robust mixtures in the presence of measurement errors

Jianyong Sun·Ata Kaban·Somak Raychaudhury
View original source ↗Read source document ↗Export metadata ↓

Abstract

We develop a mixture-based approach to robust density modeling and outlier detection for experimental multivariate data that includes measurement error information. Our model is designed to infer atypical measurements that are not due to errors, aiming to retrieve potentially interesting peculiar objects. Since exact inference is not possible in this model, we develop a tree-structured variational EM solution. This compares favorably against a fully factorial approximation scheme, approaching the accuracy of a Markov-Chain-EM, while maintaining computational simplicity. We demonstrate the benefits of including measurement errors in the model, in terms of improved outlier detection rates in varying measurement uncertainty conditions. We then use this approach in detecting peculiar quasars from an astrophysical survey, given photometric measurements with errors.

Explore related subjects

astro-ph

Keep this discovery

BibTeXRIS

Jianyong Sun, Ata Kaban, Somak Raychaudhury. 2007-09-06. Robust mixtures in the presence of measurement errors. https://arxiv.org/abs/0709.0928

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2015

Solutions to the "General Grand Unification Problem," and the Questions "How Did Our Universe Come Into Being?" and "Of What is Empty Space Composed?"

Using mathematical techniques to model one of the most simplistic of human linguistic processes, it is rationally predicted that within the nonstandard physical world (NSP-world) there exists a force-like (logical) operator *S and an entity w' such that *S{w'} sequentially generates each of the Natural systems that comprise a Universe. This model shows specifically that within the NSP-world the behavior of each Natural world Natural system is related logically. Further, the model predicts the rational existence of a single type of entity within the NSP-world's substratum that can be used to construct, by means of an exceptionally simple process, all of the fundamental Natural world particles used within particle physics. In section 11.2, it is shown how (Natural law) allowable perturbations in Natural system behavior are also included within this mathematical model. These results solve the pre-geometry problem of Wheeler. In general, the model predicts that when the behavior of these Universe creating processes is viewed globally, it can be described as apparently mirroring the behavior of an infinitely powerful computer or mind.

astro-ph↗
arXiv2014

Dark matter in elliptical galaxies: I. Is the total mass density profile of the NFW form or even steeper?

Elliptical galaxies are modelled as Sersic luminosity distributions with density profiles (DPs) for the total mass adopted from the DPs of haloes within dissipationless LambdaCDM N-body simulations. Ellipticals turn out to be inconsistent with cuspy low-concentration NFW models representing the total mass, nor are they consistent with a steeper -1.5 inner slope, nor with the shallower models proposed by Navarro et al. 04, nor with NFW models 10 times more concentrated than predicted, as deduced from several X-ray observations: the mass models, extrapolated inwards, lead to local mass-to-light ratios that are smaller than the stellar value inside an effective radius (R_e), and to central aperture velocity dispersions that are much smaller than observed. This conclusion remains true as long as there is no sharp steepening (slope < -2) of the dark matter (DM) DPs just inside 0.01 virial radii. The too low total mass and velocity dispersion produced within R_e by an NFW-like total mass profile suggests that the stellar component should dominate the DM one out to at least R_e. It should then be difficult to kinematically constrain the inner slope of the dark matter DP of ellipticals. The high concentration parameters deduced from X-ray observations appear to be a consequence of fitting an NFW model to the total mass DP made up of a stellar component that dominates inside and a DM component that dominates outwards. An appendix gives the virial mass dependence of the concentration parameter, central density, and total mass of the Navarro et al. model. In a 2nd appendix are given single integral expressions for the velocity dispersions averaged along the line-of-sight, in circular apertures and in thin slits, for general luminosity density and mass distributions, with isotropic orbits.

astro-ph↗
arXiv2014

Dark matter in elliptical galaxies: II. Estimating the mass within the virial radius

Elliptical galaxies are modelled with a a 4-component model: Sersic stars, LCDM dark matter (DM), hot gas and central black hole. DM is negligible in the inner regions, which are dominated by stars and the central black hole. This prevents any kinematical estimate (using a Jeans analysis) of the inner slope of the DM density profile. The gas fraction rises, but the baryon fraction decreases with radius, at least out to 10 effective radii (R_e). Even with line-of-sight velocity dispersion (VD) measurements at 4 to 6 R_e with 20 km/s accuracy and perfectly known velocity anisotropy, the total mass within the virial radius (r_v) is uncertain by a factor over 3. The DM distributions found in LCDM simulations are consistent with the stellar VD profiles, but appear inconsistent with the low VDs measured by Romanowsky et al. (2003) of planetary nebulae between 2 and 5 R_e, which imply such low M/Ls that the baryon fraction within r_v must be greater than the universal value. Replacing the NFW DM model by the new model of Navarro et al. (2004) decreases slightly the VD at a given radius. So, given the observed VD measured at 5 R_e, the inferred M/L within r_v is 40% larger than predicted with the NFW model. Folding in the slight (strong) radial anisotropy found in LCDM (merger) simulations, which is well modelled (much better than with the Osipkov-Merritt formula) with beta(r) = 1/2 r/(r+a), the inferred M/L within r_v is another 1.6 (2.4) times higher than for the isotropic NFW model. Thus, the DM model and radial anisotropy can partly explain the low PN VDs, but not in full. In an appendix, single integral expressions are derived for the VDs in terms of the tracer density and total mass profiles, for 3 anisotropic models: radial, Osipkov-Merritt, and the model above, for general radial profiles of luminosity density and mass.

astro-ph↗