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:20260114T163646Z
LOCATION:Meeting Room C4.8\, Level 4 (Convention Centre)
DTSTART;TZID=Australia/Melbourne:20231215T160500
DTEND;TZID=Australia/Melbourne:20231215T161500
UID:siggraphasia_SIGGRAPH Asia 2023_sess159_papers_397@linklings.com
SUMMARY:Compact Neural Graphic Primitives with Learned Hash Probing
DESCRIPTION:Towaki Takikawa (NVIDIA, University of Toronto); Thomas Müller
 , Merlin Nimier-David, and Alex Evans (NVIDIA); Sanja Fidler (NVIDIA, Univ
 ersity of Toronto); Alec Jacobson (University of Toronto, Adobe Research);
  and Alexander Keller (NVIDIA)\n\nNeural graphics primitives are faster an
 d achieve higher quality when their neural networks are augmented by spati
 al data structures that hold trainable features arranged in a grid. Howeve
 r, existing feature grids either come with a large memory footprint (dense
  or factorized grids, trees, and hash tables) or slow performance (index l
 earning and vector quantization). In this paper, we show that a hash table
  with learned probes has neither disadvantage, resulting in a favorable co
 mbination of size and speed. Inference is faster than unprobed hash tables
  at equal quality while training is only 1.2-2.6x slower, significantly ou
 tperforming prior index learning approaches. We arrive at this formulation
  by casting all feature grids into a common framework: they each correspon
 d to a lookup function that indexes into a table of feature vectors. In th
 is framework, the lookup functions of existing data structures can be comb
 ined by simple arithmetic combinations of their indices, resulting in Pare
 to optimal compression and speed.\n\nRegistration Category: Full Access\n\
 nSession Chair: Fei Hou (Institute of Software, Chinese Academy of Science
 s; University of Chinese Academy of Sciences)\n\n
URL:https://asia.siggraph.org/2023/full-program?id=papers_397&sess=sess159
END:VEVENT
END:VCALENDAR