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:20250110T023313Z
LOCATION:Hall B7 (1)\, B Block\, Level 7
DTSTART;TZID=Asia/Tokyo:20241206T093400
DTEND;TZID=Asia/Tokyo:20241206T094600
UID:siggraphasia_SIGGRAPH Asia 2024_sess141_papers_776@linklings.com
SUMMARY:Accelerate Neural Subspace-Based Reduced-Order Solver of Deformabl
 e Simulation by Lipschitz Optimization
DESCRIPTION:Technical Papers\n\nAoran Lyu (South China University of Techn
 ology, University of Manchester); Shixian Zhao, Chuhua Xian, Zhihao Cen, a
 nd Hongmin Cai (South China University of Technology); and Guoxin Fang (Ch
 inese University of Hong Kong)\n\nReduced-order simulation is an emerging 
 method for accelerating physical simulations with high DOFs, and recently 
 developed neural-network-based methods with nonlinear subspaces have been 
 proven effective in diverse applications as more concise subspaces can be 
 detected. However, the complexity and landscape of simulation objectives w
 ithin the subspace have not been optimized, which leaves room for enhancem
 ent of the convergence speed. This work focuses on this point by proposing
  a general method for finding optimized subspace mappings, enabling furthe
 r acceleration of neural reduced-order simulations while capturing compreh
 ensive representations of the configuration manifolds. We achieve this by 
 optimizing the Lipschitz energy of the elasticity term in the simulation o
 bjective, and incorporating the cubature approximation into the training p
 rocess to manage the high memory and time demands associated with optimizi
 ng the newly introduced energy. Our method is versatile and applicable to 
 both supervised and unsupervised settings for optimizing the parameterizat
 ions of the configuration manifolds. We demonstrate the effectiveness of o
 ur approach through general cases in both quasi-static and dynamics simula
 tions. Our method achieves acceleration factors of up to 6.83 while consis
 tently preserving comparable simulation accuracy in various cases, includi
 ng large twisting, bending, and rotational deformations with collision han
 dling. This novel approach offers significant potential for accelerating p
 hysical simulations, and can be a good add-on to existing neural-network-b
 ased solutions in modeling complex deformable objects.\n\nRegistration Cat
 egory: Full Access, Full Access Supporter\n\nLanguage Format: English Lang
 uage\n\nSession Chair: Sheldon Andrews (École de technologie supérieure (É
 TS))
URL:https://asia.siggraph.org/2024/program/?id=papers_776&sess=sess141
END:VEVENT
END:VCALENDAR