Automated%20Classification%20of%20X-ray%20Sources

1
Automated Classificationof X-ray Sources

• R. J. Hanisch, A. A. Suchkov, R. L. White
• Space Telescope Science Institute
•  
• T. A. McGlynn, E. L. Winter, M. F. Corcoran
• NASA Goddard Space Flight Center
•  
• W. Voges
• Max-Planck-Institute for Extraterrestrial Physics

Supported by NASAs Applied Information Systems
Research Program, Grant NAG5-11019
2
ClassX

• ClassX is a Virtual Observatory prototype project
  aimed at the semi-automated classification of
  unidentified X-ray sources.
• ClassX draws from numerous on-line object
  catalogs using VO standard protocols (cone
  search, VOTable) to collect multi-wavelength
  position, flux, and source extent information.
• ClassX uses these data to train oblique decision
  tree classifiers, and then apply the classifiers
  to unidentified X-ray sources.

3
ClassX Overview
Use existing high-coverage resources to get
information on user sources.
Initially use small-coverage resources for
verification.
WGACAT
SDSS
RASS
1
HST
USNOA2
Chandra
DSS
XMM
GSC2
2
3
2MASS
NVSS
FIRST
4
What Kind of Classifier?
Classifiers can be distinguished along several
orthogonal dimensions. Exploring all the
dimensions is hard. Different tasks may require
different classifiers.
Classifier algorithm Decision trees, oblique or
otherwise Neural networks Nearest neighbor
Observed quantities Fluxes, positions, colors,
variability, spatial extent, X-ray, optical,
IR, ...
Classification granularity Coarse Stellar vs.
Extragalactic Fine A0 vs. B0, AGN vs. QSO vs.
galaxy
Training sets WGACAT, ROSAT All Sky Survey, ...
5
ClassX Performance
X-ray, opt.
X-ray, opt., IR

• Every output class needs substantial
  representation in the training set.
• Overlap between classes should be minimized.
• Classifier accuracy can be improved with
  additional information (i.e., flux in different
  bandpass), but not always!
• Stellar and extragalactic sources are easily
  distinguished.

Almost no extra- galactic sources classified as
stars
Small amount of confusion among different stellar
types
Very few stars classified as extragalactic sources
Extragalactic source classifications
more ambiguous owing to class overlap
6
X-ray Stars in ? Oph

• 10X increase in number of identified X-ray stars
• Dominance of late-type stars consistent with
  large pre-main-sequence population in active star
  formation region ? T Tauri-type stars
• Adds many new T Tauri candidates

7
X-ray Stars in the LMC

• 10X increase in number of identified early-type
  stars
• Dominance of early-type stars is consistent with
  expectations for stars at distance of LMC
• Many late-type X-ray stars suggest large
  population of PMS T Tauri stars in LMC star
  formation regions

8
X-ray Binaries
Additional XRBs?
X-ray hardness ratio
mx1 (soft x-ray magnitude)

• WGACAT stars (type unknown) re-classified most
  are indeed stars, most in direction of LMC/SMC
• 53 new XRB candidates 50 increase in number
  known in WGACAT. These are mostly high-mass XRB
  candidates with bright optical counterparts.

9
Quasars and AGN

• Nearly 20X increase in number of QSO candidates,
  3X increase in number of AGNs. ClassX
  differentiates reasonably well between QSOs and
  AGN.
• In contrast to QSO/AGN objects known in WGACAT,
  where dominant class is AGN, objects identified
  by ClassX are strongly dominated by QSOs. On
  average are much fainter in the X-rays, by more
  than 1 mag also substantially redder in the
  optical.
• Of ClassX-classified QSOs in region of SDSS EDR,
  60 are confirmed.

Known sources
ClassX sources
10
Summary

• Core technology of ClassX in place and working
  effectively.
• Suite of classifiers developed.
• Initial results in areas of stellar X-ray
  sources
• Pre-main-sequence stars, T Tauri stars readily
  identified in galactic star formation regions
• Large increase in numbers of both early- and
  late-type X-ray stars in LMC
• 50 increase in number of candidate X-ray
  binaries
• and quasars/AGN
• Identifying faint, high-redshift QSOs
• Pursuing further validation, e.g., through SDSS,
  HST, and Chandra observations
• http//heasarc.gsfc.nasa.gov/classx/