BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:On Low Dimensional Random Projections and Similarity Search - Lu\,
  Yu-En (Eric) (Univeristy of Cambridge)
DTSTART:20081010T130000Z
DTEND:20081010T140000Z
UID:TALK13295@talks.cam.ac.uk
CONTACT:Eiko Yoneki
DESCRIPTION:Random projection (RP) is a common technique for dimensionalit
 y reduction under $L_2$ norm for which many significant space embedding re
 sults have been demonstrated.\nHowever\, many similarity search applicatio
 ns often require very low dimension embeddings in order to reduce overhead
  and boost performance.\nFor example\, a good 1D embedding can enable comp
 lex queries over standard distributed hash tables.\n\nInspired by the use 
 of symmetric probability distributions in previous work\, we propose a nov
 el RP algorithm\, Beta Random Projection\, and give its probabilistic anal
 yses based on Beta and Gaussian approximations. We evaluate the algorithm 
 in terms of standard similarity metrics with other RP algorithms as well a
 s the singular value decomposition (SVD). Our experimental results show th
 at BRP preserves both similarity metrics well and\, under various dataset 
 types including random point sets\, text (TREC5) and images\, provides sha
 rper and consistent performance.\n
LOCATION:SS03\, Computer Laboratory\, William Gates Builiding
END:VEVENT
END:VCALENDAR