Linde–Buzo–Gray algorithm: Difference between revisions

Revision as of 07:57, 28 December 2023

Latest revision as of 00:44, 31 July 2025

(6 intermediate revisions by 4 users not shown)

Line 1:

{{missing information|general information, usage in the field (mention cinepak?), optimality conditions, choice of {{epsilon}}s, model instead of training data, ELBG|date=December 2023}}

{{one source|date=December 2023}}

The '''Linde–Buzo–Gray algorithm''' (named after its creators Yoseph Linde, Andrés Buzo and [[Robert M. Gray]], who designed it in 1980)<ref>{{Cite journal| doi = 10.1109/TCOM.1980.1094577| issn = 0090-6778| volume = 28| issue = 1| pages = 84–95| last1 = Linde| first1 = Y.| last2 = Buzo| first2 = A.| last3 = Gray| first3 = R.| title = An Algorithm for Vector Quantizer Design| journal = IEEE Transactions on Communications| access-date = 2023-12-28| date = 1980| urls2cid = https://ieeexplore.ieee.org/document/1094577/18530691}}</ref> is an [[Iterative method|iterative]] [[vector quantization]] algorithm to improve a small set of vectors (codebook) to represent a larger set of vectors (training set), such that it will be [[Local optimum|locally optimal]]. It combines [[Lloyd's Algorithm]] with a splitting technique in which larger codebooks are built from smaller codebooks by splitting each code vector in two with [[Lloyd's Algorithm]]. The core idea of the algorithm is that by splitting the codebook such that all code vectors from the previous codebook are present, the new codebook must be as good as the previous one or better. <ref name=gray1992>{{Cite book| edition = 1| publisher = Springer| isbn = 978-1-4613-6612-6| last1 = Gray| first1 = R.| last2 = Gersho| first2 = A.| title = Vector Quantization and Signal Compression| date = 1992| doi = 10.1007/978-1-4615-3626-0| url = https://doi.org/10.1007/978-1-4615-3626-0}}</ref>{{rp|361-362}}

== Description ==

Line 11:

'''output''': codebook that is twice the size and better or as good as ''old-codebook''

''new-codebook'' &larr;← {}

'''for each''' ''old-codevector'' '''in''' ''old-codebook'' '''do'''

Line 24:

'''do'''

''previous-codebook'' &larr;← ''codebook''

''clusters'' &larr;← divide ''training'' into |''codebook''| clusters, where each cluster contains all vectors in ''training'' who are best represented by the corresponding vector in ''codebook''

'''for each''' cluster ''cluster'' in ''clusters'' '''do'''

the corresponding code vector in ''codebook'' &larr;← the centroid of all training vectors in ''cluster''

'''while''' the difference in distortionerror representing ''training'' between ''codebook'' and ''previous-codebook'' &gt; {{epsilon}}

'''return''' ''codebook''