BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Europe/Stockholm
X-LIC-LOCATION:Europe/Stockholm
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20190719T085744Z
LOCATION:HG EO Nord
DTSTART;TZID=Europe/Stockholm:20190613T195000
DTEND;TZID=Europe/Stockholm:20190613T215000
UID:submissions.pasc-conference.org_PASC19_sess179_post145@linklings.com
SUMMARY:PHY03 - Comparison of Machine Learning Techniques for Emulating Co
 llisions in Planet Formation
DESCRIPTION:Poster\n\n\nPHY03 - Comparison of Machine Learning Techniques 
 for Emulating Collisions in Planet Formation\n\nTimpe, Knabenhans, Han Vei
 ga\n\nCollisions between growing bodies are the fundamental agent of plane
 t formation. However, despite their importance, the computational cost of 
 simulating each collision, coupled with the high-dimensionality of the inp
 ut parameter space and non-linear response of the output parameter space, 
 has so far frustrated a comprehensive understanding of collisions. Indeed,
  only limited regions of the input parameter space have been explored, wit
 h most simulations focused on specific problems in planetary science, such
  as the origin of Earth’s moon and Mercury’s large core. Follo
 wing the success of emulation techniques applied to large-scale structure 
 formation in cosmology and recent attempts to classify and predict collisi
 on outcomes in planet formation, we investigate the ability of different u
 ncertainty quantification and machine learning techniques to accurately em
 ulate pairwise collisions. In particular, we focus on the complexity, perf
 ormance, dataset requirements, and other measures which indicate the most 
 appropriate techniques for this particular type of problem. This work is b
 ased on a set of 10,000 simulations of pairwise collisions, which is an or
 der-of-magnitude larger than any previous dataset used in similar studies.
  This dataset allows us to probe regions of the input parameter space neve
 r before explored.
END:VEVENT
END:VCALENDAR