An automated, rapid classification of transient events detected in the modernsynoptic sky surveys is essential for their scientific utility and effectivefollow-up using scarce resources. This presents some unusual challenges: thedata are sparse, heterogeneous and incomplete; evolving in time; and most ofthe relevant information comes not from the data stream itself, but from avariety of archival data and contextual information (spatial, temporal, andmulti-wavelength). We are exploring a variety of novel techniques, mostlyBayesian, to respond to these challenges, using the ongoing CRTS sky survey asa testbed. The current surveys are already overwhelming our ability toeffectively follow all of the potentially interesting events, and thesechallenges will grow by orders of magnitude over the next decade as the moreambitious sky surveys get under way. While we focus on an application in aspecific domain (astrophysics), these challenges are more broadly relevant forevent or anomaly detection and knowledge discovery in massive data streams.
展开▼
