BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Toward Foundation Models for Seismology and Geophysics - Maarten d
 e Hoop (Rice University)
DTSTART:20260401T130000Z
DTEND:20260401T140000Z
UID:TALK246211@talks.cam.ac.uk
CONTACT:Sergei Lebedev
DESCRIPTION:Deep learning has rapidly transformed seismology\, and more ge
 nerally geophysics\, by shifting the focus from task-specific algorithms t
 o learning general representations directly from data. Early successes cam
 e from supervised applications such as earthquake\, as well as polyphonic 
 seismo-volcanic signals detection upon introducing the "scattering network
 ”\, and phase picking but a key advance occurred with unsupervised (deep
  clustering) and self-supervised methods which we developed and that can o
 rganize seismic waveforms without reliance on labeled catalogs. These appr
 oaches uncover latent structure across multiple time scales and have enabl
 ed the discovery (and separation) of previously undetected events and subt
 le seismic phenomena. Building on these results\, we introduced SeisLM\, a
  large-scale pretrained (transformer-based) "foundation model"\, trained o
 n continuous global datasets to provide transferable representations that 
 can be adapted across regions and tasks through fine tuning with minimal a
 dditional data\, such as tremor detection associated with slow slip events
 . We show how self attention mechanisms naturally support forecasting\, an
 d introduce HARPA\, a high-rate phase association framework that lifts arr
 ival sequences associated with (unknown) microseismic events from arrays t
 o probability distributions and compares them using an optimal transport m
 etric\; a generative travel time neural field is used to estimate the wave
  speed (and event locations) simultaneously with association. Leveraging a
 n architecture reminiscent of cross attention encoding geometry\, we furth
 er demonstrate that ray transforms - and implicitly the underlying wave sp
 eeds - can be learned directly from travel-time data. Finally\, we present
  designs of foundation models\, including Flowers\, combining data-driven 
 with physical principles\, aimed at enabling scientific discovery. We show
  examples of approximating wave propagation and scattering\, and fluid dyn
 amics through data-driven surrogates.
LOCATION:Wolfson Lecture Theatre
END:VEVENT
END:VCALENDAR