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 F 3
DTSTART;TZID=Europe/Stockholm:20190613T114500
DTEND;TZID=Europe/Stockholm:20190613T121500
UID:submissions.pasc-conference.org_PASC19_sess120_msa261@linklings.com
SUMMARY:Neural Concept Shape and Some Applications: a Deep-Learning Framew
 ork for CAE
DESCRIPTION:Minisymposium\nComputer Science and Applied Mathematics, Emerg
 ing Application Domains, Chemistry and Materials, Climate and Weather, Phy
 sics, Solid Earth Dynamics, Life Sciences, Engineering\n\nNeural Concept S
 hape and Some Applications: a Deep-Learning Framework for CAE\n\nBaque\n\n
 Shape optimization through simulations has many industrial applications. E
 xisting methods, however, are so computationally demanding that typical en
 gineering practices are to either simply try a limited number of hand-desi
 gned shapes or restrict oneself to shapes that can be parameterized using 
 only few degrees of freedom. Neural Concept Shape is a new software packag
 e that introduces a novel way to optimize complex shapes fast and accurate
 ly through surrogate modeling. To this end, we train Geodesic Convolutiona
 l Neural Networks to emulate a simulator directly using a CAD representati
 on of the input. The neural net is then used to formulate an objective fun
 ction that is differentiable with respect to the shape parameters, which c
 an then be optimized using a gradient-based technique. This outperforms st
 ate-of-the-art methods by 5 to 20\% for standard problems and, even more i
 mportantly, this approach applies to cases that previous methods cann
 ot handle.
END:VEVENT
END:VCALENDAR