MACHO QSO Candidates in the LMC Fields:

Selected Using Machine Learning
Trained on Time Variability and Multiple Diagnostics

  This webpage provides the catalogs and lightcurves (MACHO B and R bands) of 1) the known 58 MACHO QSOs and 2) the 2,566 QSO candidates in the LMC fields. In these catalogs, we complied number of properties of the objects including RA, Dec, crossmatched IDs with several catalogs, magnitudes, spectroscopic redshifts for the 58 MACHO QSOs, photometric redshifts, etc.

  See Kim et al 2011a for the SVM (a.k.a Support Vector Machine) QSO classification model based on the time variability of lightcurves. We used the model to select the 2,566 QSO candidates.

  In our recent work (Kim et al. 2011b), we employed multiple diagnostics, such as X-ray flux, mid-IR color and AGN SED fitting, in order to select 663 promising QSO candidates among the 2,566 candidates. These candidates are flagged in the catalog. See Figure 1 shown below for the comparison of efficiency of Kim et al 2011a and Kim et al. 2011b.

  Note that we calibrated MACHO RA and Dec of the candidates using the UCAC3 catalog and improved the average astrometric accuracy from ~3'' to ~0.5''.

known MACHO QSOs
QSO candidates

  We provide two types of catalogs. One is an ascii-type catalog and another is a fits-type catalog. We recommend to use fits-type catalogs, which also provides comments for some columns. You can use any software that can read fits-type catalogs such as TopCat. Check ReadMe file which explains the columns.

Figure 1. Comparison of efficiency between the previous (Kim et al 2011a) and current (Kim et al. 2011b) method. As the figure shows, the current method is superior to the previous method.

Any question and/or suggestion to :

We used Python programming language for these projects. The analysis in these papers have been done using the Odyssey cluster supported by the FAS Research Computing Group at Harvard.

@ last modified at 01/04/2012