BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:The ATLAS Virtual Research Assistant - Heloise Stevance\, Universi
 ty of Oxford
DTSTART:20250210T120000Z
DTEND:20250210T123000Z
UID:TALK227629@talks.cam.ac.uk
CONTACT:Sam Nallaperuma-Herzberg
DESCRIPTION:The ATLAS sky survey is able to image the whole night sky ever
 y 24 to 48 hours\, looking for near earth asteroids and exploding stars. \
 nThis generates 10s of millions of potential alerts every day which must b
 e triaged and filtered to find the few explosions worth following up with 
 more (expensive and time consuming) resources.\nMuch of the work can be au
 tomated\, but human "eyeballing" remains the final step before reporting a
 nd follow-up.\nThis is a task that involves crappy images\, sparse and une
 ven time series (with error bars and non-detections)\, and a whole lot of 
 contextual knowledge to make sense of the mess. \nWe are looking for rare 
 events (not many training samples)\, we want (near) 100% recall\, and expl
 ainability is paramount (since faults in the VRA can impact astrophysical 
 event rates). That is a lot to ask of any model.\nIn this presentation I w
 ill briefly present how ML is used in the VRA to lower eyeballer workload 
 (decreased by 70%) and why "fancier" ML methods reported in the literature
  did not address our problems.\nI will also touch on upcoming sky survey c
 hallenges.
LOCATION:FW11\, Willam Gates building (Department of Computer Science and 
 Technology)
END:VEVENT
END:VCALENDAR