Place: 4421 Sterling Hall, Coffee and cookies 3:30 PM, Talk Begins 3:45 PM
Speaker: Lucianne Walkowicz, Chicago Planetarium
Abstract: As of last year, the Large Synoptic Survey Telescope (LSST) has begun construction on the summit of Cerro Pachon. As the top-rated flagship for ground-based astronomy in the next decade, LSST will provide an unprecedented dataset of 37 billion objects observed in both space and time. The time domain aspect of LSST is an especially promising source of new discoveries: the main survey is expected to generate new samples of thousands of supernovae, cataclysmic variables, stellar flares, and regular variables, amongst other denizens of the time-domain zoo, each one of which will generate an "alert" within 60 seconds of observation. Sorting amongst these transient and variable objects poses a challenging task: transient events of interest must be identified and prioritized, so that valuable follow-up resources (which are easily saturated by the volume of LSST alerts per night) are deployed on the events with the most potential to provide transformative understanding of particular phenomena. For LSST, this task is of course at a beyond-human scale, requiring sophisticated machine learning algorithms to provide real-time characterization and prioritization. However, another challenge looms under the surface of the approaching flood of data: how can truly novel phenomena be recognized and discovered in large datasets? In this talk, I will discuss methods and applications of finding anomalous data in astronomical datasets. Anomaly identification is a powerful means to both discover novel phenomena, as well as to identify problematic data so that it may be cleaned from the database. Lastly, hunting down anomalies is an exciting way to engage citizen scientists in astronomical discovery, whose efforts have repeatedly demonstrated the power of the crowd in uncovering previously-unnoticed phenomena.