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:20190719T085745Z
LOCATION:HG D 1.2
DTSTART;TZID=Europe/Stockholm:20190614T140000
DTEND;TZID=Europe/Stockholm:20190614T143000
UID:submissions.pasc-conference.org_PASC19_sess114_msa280@linklings.com
SUMMARY:Neuronal Network Integration in RAPS
DESCRIPTION:Minisymposium\nClimate and Weather\n\nNeuronal Network Integra
 tion in RAPS\n\nProgsch\n\nIntegrating recent advances in machine learning
  such as deep neuronal networks into existing scientific software poses un
 ique challenges. We present our experiences and results from ongoing work 
 on introducing neuronal networks into the radiation solver of RAPS. To thi
 s end we added facilities to call the machine learning framework MXnet fro
 m within the existing radiation solver infrastructure as well as the abili
 ty to extract radiation samples from a running simulation. Neuronal networ
 ks are trained using these extracted samples and can then be run in place 
 of the original radiation solver. Covered topics include challenges of int
 egrating machine learning frameworks with Fortran code, design of neuronal
  networks suited to the problem, data representations suited for machine l
 earning and debugging of neuronal networks in the context of a simulation.
END:VEVENT
END:VCALENDAR