BEGIN:VCALENDAR
VERSION:2.0
CALSCALE:GREGORIAN
PRODID:UW-Madison-Physics-Events
BEGIN:VEVENT
SEQUENCE:1
UID:UW-Physics-Event-6599
DTSTART:20210922T160000Z
DTEND:20210922T171500Z
DTSTAMP:20260414T191917Z
LAST-MODIFIED:20210908T045811Z
LOCATION:Online Seminar: Please sign up for our mailing list at www.ph
 ysicsmeetsml.org for zoom link
SUMMARY:Differentiable Physics Simulations for Deep Learning\, Physics
  ∩ ML Seminar\, Nils Thuerey\, TU Munich
DESCRIPTION:In this talk I will focus on the possibilities that arise 
 from recent advances in the area of deep learning for physical simulat
 ions. In this context\, especially the Navier-Stokes equations represe
 nt an interesting and challenging advection-diffusion PDE that poses a
  variety of challenges for deep learning methods.
\n
\nIn particular\,
  I will focus on differentiable physics solvers from the larger field 
 of differentiable programming. Differentiable solvers are very powerfu
 l tools to integrate into deep learning processes. The existing numeri
 cal methods for efficient solvers can be leveraged within learning tas
 ks to provide crucial information in the form of reliable gradients to
  update the weights of a neural networks. Interestingly\, it turns out
  to be beneficial to combine supervised and physics-based approaches. 
 The former poses a much simpler learning task by providing explicit re
 ference data that is typically pre-computed. Physics-based learning on
  the other hand can provide gradients for a larger space of states tha
 t are only encountered at training time. Here\, differentiable solvers
  are particularly powerful to\, e.g.\, provide neural networks with fe
 edback about how inferred solutions influence the long-term behavior o
 f a physical model.
\n
\nI will demonstrate this concept with several 
 examples from learning to reduce numerical errors\, over long-term pla
 nning and control\, to generalization. I will conclude by discussing c
 urrent limitations and by giving an outlook about promising future dir
 ections.
URL:https://www.physics.wisc.edu/events/?id=6599
END:VEVENT
END:VCALENDAR