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 E 1.2
DTSTART;TZID=Europe/Stockholm:20190613T154500
DTEND;TZID=Europe/Stockholm:20190613T161500
UID:submissions.pasc-conference.org_PASC19_sess135_msa311@linklings.com
SUMMARY:Exploiting Wavefield Complexity, Multi-Physics and Deep Learning f
 or Seismic Modeling and Inversion
DESCRIPTION:Minisymposium\nComputer Science and Applied Mathematics, Physi
 cs, Solid Earth Dynamics\n\nExploiting Wavefield Complexity, Multi-Physics
  and Deep Learning for Seismic Modeling and Inversion\n\nNissen-Meyer, Len
 g, Moseley, Fernando, Haindl...\n\nDespite significant progress in algorit
 hmic and hardware developments, the seismic wave propagation and inversion
  problem is still daunting even on largest HPC infrastructures. In this ta
 lk, we explore various directions to harness physics-based and algorithmic
  advance to reach novel parameter regimes, including wave propagation in g
 lobal ocean layers and – for the first time – the maximal seis
 mic frequency resolution of 1Hz for realistic 3D Earth models. This enable
 s us further to consider deterministic and diffuse scattering regimes join
 tly, leading to novel considerations on locality and spectral properties o
 f heterogeneities. We will also present an approach to solve the full
 -waveform inversion problem with these accelerated methods using discrete 
 adjoints. Most of the above is based on our novel solver AxiSEM3D, wh
 ich exploits wavefield smoothness by a learning procedure and utilizes the
  eigen library for accelerated matrix-vector computations. Next, we apply 
 various deep-learning networks to solve seismic wave propagation and 
 inversion, suggesting that such approaches enable drastic speedups compare
 d to discrete methods, especially if combined wit physical motiv
 ations and aided by existent numerical methods, at least for simpler setti
 ngs. Applications of these advances and complementary projects range from 
 global earth tomographic imaging to deciphering African elephant vibr
 ations.
END:VEVENT
END:VCALENDAR