BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Asia/Tokyo
X-LIC-LOCATION:Asia/Tokyo
BEGIN:STANDARD
TZOFFSETFROM:+0900
TZOFFSETTO:+0900
TZNAME:JST
DTSTART:18871231T000000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260817T171515Z
LOCATION:Hall C\, C Block\, Level 4
DTSTART;TZID=Asia/Tokyo:20241203T090000
DTEND;TZID=Asia/Tokyo:20241203T120000
UID:siggraphasia_SIGGRAPH Asia 2024_sess223_papers_387@linklings.com
SUMMARY:Differential Walk on Spheres
DESCRIPTION:Bailey Miller (Carnegie Mellon Uniersity), Rohan Sawhney (NVID
 IA), and Keenan Crane and Ioannis Gkioulekas (Carnegie Mellon Uniersity)\n
 \nWe introduce a Monte Carlo method for computing derivatives of the solut
 ion to a partial differential equation (PDE) with respect to problem param
 eters (such as domain geometry or boundary conditions). Derivatives can be
  evaluated at arbitrary points, without performing a global solve or const
 ructing a volumetric grid or mesh. The method is hence well suited to inve
 rse problems with complex geometry, such as PDE-constrained shape optimiza
 tion. Like other walk on spheres (WoS) algorithms, our method is trivial t
 o parallelize, and is agnostic to boundary representation (meshes, splines
 , implicit surfaces, etc.), supporting large topological changes.  We focu
 s in particular on screened Poisson equations, which model diverse problem
 s from scientific and geometric computing.  As in differentiable rendering
 , we jointly estimate derivatives with respect to all parameters---hence, 
 cost does not grow significantly with parameter count.  In practice, even 
 noisy derivative estimates exhibit fast, stable convergence for stochastic
  gradient-based optimization, as we show through examples from thermal des
 ign, shape from diffusion, and computer graphics.\n\nRegistration Category
 : Enhanced Access, Full Access, Full Access Supporter\n\nLanguage Format: 
 English Language\n\n
URL:https://asia.siggraph.org/2024/program/?id=papers_387&sess=sess106
END:VEVENT
END:VCALENDAR