How to Find More Supernovae with Less Work: Object ClassificationTechniques for Difference Imaging

• Published in: The Astrophysical journal Vol. 665; no. 2
• Main Authors: Bailey, Stephen, Aragon, Cecilia, Romano, Raquel, Thomas, RollinC, Weaver, Benjamin A., Wong, Daniel
• Format: Journal Article
• Language: English
• Published: United States 02-05-2007
• Subjects: ASTEROIDS; CLASSIFICATION; DETECTION; EFFICIENCY; FORESTS; PIPELINES; SHAPE; SUPERNOVAE; Supernovae Boosted Decision Trees ClassificationMethods; TRANSIENTS; TREES; VECTORS
• ISSN: 0004-637X; 1538-4357
• DOI: 10.1086/519832

We present the results of applying new object classificationtechniques to difference images in the context of the Nearby SupernovaFactory supernova search. Most current supernova searches subtractreference images from new images, identify objects in these differenceimages, and apply simple threshold cuts on parameters such as statisticalsignificance, shape, and motionto reject objects such as cosmic rays,asteroids, and subtraction artifacts. Although most static objectssubtract cleanly, even a very low false positive detection rate can leadto hundreds of non-supernova candidates which must be vetted by humaninspection before triggering additional followup. In comparison to simplethreshold cuts, more sophisticated methods such as Boosted DecisionTrees, Random Forests, and Support Vector Machines provide dramaticallybetter object discrimination. At the Nearby Supernova Factory, we reducedthe number of non-supernova candidates by a factor of 10 while increasingour supernova identification efficiency. Methods such as these will becrucial for maintaining a reasonable false positive rate in the automatedtransient alert pipelines of upcoming projects such as PanSTARRS andLSST.