BEGIN:VCALENDAR VERSION:2.0 PRODID:researchseminars.org CALSCALE:GREGORIAN X-WR-CALNAME:researchseminars.org BEGIN:VEVENT SUMMARY:Mariel Pettee (Lawrence Berkeley National Laboratory / Flatiron In stitute Center for Computational Astrophysics) DTSTART:20230511T163000Z DTEND:20230511T183000Z DTSTAMP:20260423T004734Z UID:nhetc/77 DESCRIPTION:Title: W eakly-Supervised Anomaly Detection in the Milky Way\nby Mariel Pettee (Lawrence Berkeley National Laboratory / Flatiron Institute Center for Com putational Astrophysics) as part of NHETC Seminar\n\n\nAbstract\nClassific ation Without Labels (CWoLa) is a weakly-supervised anomaly detection tech nique that leverages neural networks to identify cold stellar streams with in the more than one billion Milky Way stars observed by the Gaia satellit e. The CWoLa methodology operates without the use of labeled streams or kn owledge of astrophysical principles. Instead\, it uses a classifier to dis tinguish between mixed samples for which the proportions of signal and bac kground samples are unknown. This computationally lightweight strategy is able to detect both simulated streams and the known stream GD-1 in data. O riginally designed for high-energy collider physics\, this technique may h ave broad applicability within astrophysics as well as other domains inter ested in identifying localized anomalies.\n LOCATION:https://researchseminars.org/talk/nhetc/77/ END:VEVENT END:VCALENDAR