BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Australia/Melbourne
X-LIC-LOCATION:Australia/Melbourne
BEGIN:DAYLIGHT
TZOFFSETFROM:+1000
TZOFFSETTO:+1100
TZNAME:AEDT
DTSTART:19721003T020000
RRULE:FREQ=YEARLY;BYMONTH=4;BYDAY=1SU
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:19721003T020000
TZOFFSETFROM:+1100
TZOFFSETTO:+1000
TZNAME:AEST
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260114T163648Z
LOCATION:Meeting Room C4.8\, Level 4 (Convention Centre)
DTSTART;TZID=Australia/Melbourne:20231215T143000
DTEND;TZID=Australia/Melbourne:20231215T144000
UID:siggraphasia_SIGGRAPH Asia 2023_sess171_papers_333@linklings.com
SUMMARY:LiCROM: Linear-Subspace Continuous Reduced Order Modeling with Neu
 ral Fields
DESCRIPTION:Yue Chang (University of Toronto), Peter Yichen Chen (MIT CSAI
 L), Zhecheng Wang (University of Toronto), Maurizio M. Chiaramonte and Kev
 in Carlberg (Meta Reality Labs Research), and Eitan Grinspun (University o
 f Toronto)\n\nLinear reduced-order modeling (ROM) simplifies complex simul
 ations by approximating the behavior of a system using a simplified kinema
 tic representation. Typically, ROM\nis trained on input simulations create
 d with a specific spatial discretization, \nand then serves to accelerate 
 simulations with the same discretization. \nThis discretization-dependence
  is restrictive. \n\nBecoming independent of a specific discretization wou
 ld provide flexibility to mix and match  mesh resolutions, connectivity, a
 nd type (tetrahedral, hexahedral) in training data; to \naccelerate simula
 tions with novel discretizations unseen during training; \nand to accelera
 te adaptive simulations that temporally or parametrically change\nthe disc
 retization. \n\nWe present a flexible, discretization-independent approach
  to reduced-order modeling. \nLike traditional ROM, we represent the confi
 guration as a linear combination of displacement\nfields. Unlike tradition
 al ROM, our displacement fields are continuous maps from every point on th
 e reference domain to a corresponding displacement vector; these maps are\
 nrepresented as implicit neural fields.\n\nWith linear continuous ROM (LiC
 ROM), our training set can include multiple geometries undergoing multiple
  loading conditions, independent of their discretization. This opens the d
 oor to novel applications of reduced order modeling. For instance, we can 
 accelerate\nsimulations on geometries unseen during training, and simulati
 ons that modify the geometry at runtime, for instance via cutting, hole pu
 nching, and even swapping the entire mesh. Indeed, we achieve one-shot gen
 eralization by training on a single geometry but testing on multiple unsee
 n geometries.\n\nRegistration Category: Full Access\n\nSession Chair: Qixi
 ng Huang (University of Texas Austin)\n\n
URL:https://asia.siggraph.org/2023/full-program?id=papers_333&sess=sess171
END:VEVENT
END:VCALENDAR