User: Guest  Login
Title:

Compression for quadratic similarity queries via shape-gain quantizers

Document type:
Konferenzbeitrag
Author(s):
Dempfle, S.; Steiner, F.; Ingber, A.; Weissman, T.
Abstract:
We study the problem of compression of a Gaussian vector for the purpose of similarity identification, where similarity is defined by the mean square Euclidean distance between vectors. While the asymptotical fundamental limits of the problem - the minimal compression rate and the error exponent - were found in a previous work, in this paper we focus on the nonasymptotic domain. We first present a finite blocklength achievability bound based on shape-gain quantization: The gain (amplitude) of th...     »
Keywords:
asymptotical fundamental limits,compression rate,finite blocklength achievability bound,Gaussian processes,Gaussian vector,information theory,Lattices,Leech lattice,mean square Euclidean distance,nonasymptotic domain,quadratic similarity queries,quantisation (signal),Quantization (signal),query processing,scalar quantization,Shape,shape-gain quantization,shape-gain quantizers,similarity identification,Tin,Upper bound,Vectors,wrapped spherical code
Book / Congress title:
IEEE International Symposium on Information Theory
Year:
2014
Pages:
2839--2843
Fulltext / DOI:
doi:10.1109/ISIT.2014.6875352
 BibTeX