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:20240214T070245Z
LOCATION:Meeting Room C4.8\, Level 4 (Convention Centre)
DTSTART;TZID=Australia/Melbourne:20231213T154500
DTEND;TZID=Australia/Melbourne:20231213T155500
UID:siggraphasia_SIGGRAPH Asia 2023_sess145_papers_194@linklings.com
SUMMARY:Neural Cache for Monte Carlo Partial Differential Equation Solver
DESCRIPTION:Technical Papers\n\nZilu Li (Cornell University); Guandao Yang
  (Cornell University, Stanford University); and Xi Deng, Christopher De Sa
 , Bharath Hariharan, and Steve Marschner (Cornell University)\n\nThis pape
 r presents a method that uses neural networks as a caching mechanism to re
 duce the variance of Monte Carlo Partial Differential Equation solvers, su
 ch as the Walk-on-Spheres algorithm. While these Monte Carlo PDE solvers h
 ave the merits of being unbiased and discretization-free, their high varia
 nce often hinders real-time applications. On the other hand, neural networ
 ks can approximate the PDE solution, and evaluating these networks at infe
 rence time can be very fast. However, neural-network-based solutions may s
 uffer from convergence difficulties and high bias. Our hybrid system aims 
 to combine these two potentially complementary solutions by training a neu
 ral field to approximate the PDE solution using supervision from a WoS sol
 ver. This neural field is then used as a cache in the WoS solver to reduce
  variance during inference. We demonstrate that our neural field training 
 procedure is better than the commonly used self-supervised objectives in t
 he literature. We also show that our hybrid solver exhibits lower variance
  than WoS with the same computational budget: it is significantly better f
 or small compute budgets and provides smaller improvements for larger budg
 ets, reaching the same performance as WoS in the limit.\n\nRegistration Ca
 tegory: Full Access\n\nSession Chair: Young J. Kim (Ewha Womans University
 )
URL:https://asia.siggraph.org/2023/full-program?id=papers_194&sess=sess145
END:VEVENT
END:VCALENDAR