UNIGE document Conference Proceedings
previous document  unige:94026  next document
add to browser collection
Title

Sparse ternary codes for similarity search have higher coding gain than dense binary codes

Authors
Year 2017
Abstract This paper addresses the problem of Approximate Nearest Neighbor (ANN) search in pattern recognition where feature vectors in a database are encoded as compact codes in or- der to speed-up the similarity search in large-scale databases. Considering the ANN problem from an information-theoretic perspective, we interpret it as an encoding, which maps the original feature vectors to a less entropic sparse represen- tation while requiring them to be as informative as possi- ble. We then define the coding gain for ANN search using information-theoretic measures. We next show that the clas- sical approach to this problem, which consists of binarization of the projected vectors is sub-optimal. Instead, a properly designed ternary encoding achieves higher coding gains and lower complexity.
Keywords Approximate Nearest Neighbor searchContent identificationBinary hashingCoding gainSparse representation
Full text
Proceedings (Published version) (305 Kb) - public document Free access
Structures
Research groups Computer Vision and Multimedia Laboratory
Stochastic Information Processing Group
Citation
(ISO format)
FERDOWSI, Sohrab et al. Sparse ternary codes for similarity search have higher coding gain than dense binary codes. [s.l.] : [s.n.], 2017. https://archive-ouverte.unige.ch/unige:94026

90 hits

30 downloads

Update

Deposited on : 2017-05-08

Export document
Format :
Citation style :