BEGIN:VCALENDAR
VERSION:2.0
CALSCALE:GREGORIAN
PRODID:iCalendar-Ruby
BEGIN:VEVENT
CATEGORIES:Conferences/Seminars/Lectures
DESCRIPTION:Distinguished Seminar Series in \nComputational Science and Eng
 ineering\n\nThursday October 22 at 12:00 PM ET via Zoom\n\nDifferentiable P
 hysics Simulations for Deep Learning Algorithms  \nNils Thuerey\nAssociate 
 Professor of Computer Science\nTechnical University of Munich\nhttps://ge.i
 n.tum.de/ \n \n\nAbstract:\nDifferentiable physics solvers (from the broade
 r field of differentiable programming) show particular promise for includin
 g prior knowledge into machine learning algorithms. Differentiable operator
 s were shown to be powerful tools to guide deep learning processes\, and PD
 Es provide a wide range of components to build such operators. They also re
 present a natural way for traditional solvers and deep learning methods to 
 coexist: Using PDE solvers as differentiable operators in neural networks a
 llows us to leverage existing numerical methods for efficient solvers\, e.g
 .\, to provide reliable and flexible gradients to update the weights during
  a learning run.\n\nInterestingly\, it turns out to be beneficial to combin
 e “traditional” supervised and physics-based approaches. The former poses a
  much more straightforward and more stable learning task by providing expli
 cit reference data\, while physics-based learning can provide gradients for
  a larger space of states that are only encountered at training time. Here\
 , differentiable solvers are particularly powerful\, e.g.\, to provide neur
 al networks with feedback about how inferred solutions influence a physical
  model’s long-term behavior. I will show and discuss examples with various 
 advection-diffusion type PDEs\, among others the Navier-Stokes equations fo
 r fluids\, for different learning applications. These demonstrations will h
 ighlight the properties and capabilities of PDE-powered deep neural network
 s and serve as a starting point for discussing future developments.\n\nBio:
 \nNils is currently working as Associate-Professor at the Technical Univers
 ity of Munich (TUM). He and his group focus on deep learning methods for ph
 ysical simulations\, with a particular focus on fluid phenomena. Nils acqui
 red his Ph.D. for his work on liquid simulations in 2006 from the Universit
 y of Erlangen-Nuremberg. Until 2010 he held a position as a post-doctoral r
 esearcher at ETH Zurich. He received a tech-Oscar from the AMPAS in 2013 fo
 r his research on controllable smoke effects. Subsequently\, Nils worked fo
 r three years as R&D lead at ScanlineVFX\, before starting at TUM in Octobe
 r 2013.
DTEND:20201022T170000Z
DTSTAMP:20260308T141541Z
DTSTART:20201022T160000Z
LOCATION:
SEQUENCE:0
SUMMARY:Distinguished Seminar Series in  Computational Science and Engineer
 ing
UID:tag:localist.com\,2008:EventInstance_34776368572654
URL:https://calendar.mit.edu/event/Nelson
END:VEVENT
END:VCALENDAR