Computer Science – Learning
Scientific paper
Dec 2010
adsabs.harvard.edu/cgi-bin/nph-data_query?bibcode=2010aspc..434..369p&link_type=abstract
Astronomical Data Analysis Software and Systems XIX. Proceedings of a conference held October 4-8, 2009 in Sapporo, Japan. Edite
Computer Science
Learning
Scientific paper
The statistical identification of all serendipitous X-ray sources detected by the EPIC camera is one of the tasks devoted to the Survey Science Centre (SSC) of XMM-Newton. Using a probabilistic cross-correlation of the 2XMMi catalogue with others like the SDSS DR7 or the 2MASS, we have built several samples of multiwavelength data for which various thresholds on the number of spurious associations can be applied. We create a learning sample of classified XMM sources from the SDSS spectroscopy and from the Archival Catalogue and Database Subsystem (ACDS) which is the part of the SSC pipeline that performs the cross-correlation of EPIC sources against a large collection of archival catalogues including Simbad. This allowed us to apply both supervised or unsupervised classification methods. We tested a range of classification algorithms: k-Nearest Neighbours, Mean Shifts, Kernel Density Classification, Learning Vector Quantisation, oblique decision tree (the OC1 algorithm) and Support Vector Machines. Advantages and disadvantages of each method are briefly reviewed, and their respective performances are compared. We also show an example of the application of the kernel density classification with several classes on the results of the correlation of the 2XMMi with the SDSS DR7.
Derriere Sebastien
Michel Laurent
Motch Ch.
Pineau François-Xavier
No associations
LandOfFree
Comparison of Classification Methods for XMM Sources does not yet have a rating. At this time, there are no reviews or comments for this scientific paper.
If you have personal experience with Comparison of Classification Methods for XMM Sources, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Comparison of Classification Methods for XMM Sources will most certainly appreciate the feedback.
Profile ID: LFWR-SCP-O-1384804