An Algorithm for Vector Quantizer Design

By Y. Linde, A. Buzo, Robert M. Gray. 1980

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

SoundStream: An End-to-End Neural Audio Codec (2021, methodology): “Optimal scalar quantization based on Lloyd’s algorithm [25] can be extended to a high-dimensional space via the generalized Lloyd algorithm (GLA) [26], which is very similar to k-means clustering [27].”
CAIM discretization algorithm (2004, background): “As always, a satisfactory trade-off between these two goals needs to be achieved.”
Bayes risk weighted vector quantization with posterior estimation for image compression and classification (1996, methodology): “The Lloyd or k-means clustering algorithm [11], [12], [2] for full search, unconstrained VQ involves an iterative application that consists of two 3”
One-shot learning gesture recognition from RGB-D data using bag of features (2013, background): “Although the complexity of SOMP is more expensive than VQ which has O(dMN) (Linde et al., 1980).”
The Effectiveness of Lloyd-Type Methods for the k-Means Problem (2006, methodology): “This and similar iterative descent methods soon became the dominant approaches to the problem [35, 33, 12, 31 ] (see also [19, 20, 24] and the references therein); they remain so today, and are still being improved [1, 40, 42, 28].”
Side match and overlap match vector quantizers for images (1992, methodology): “Following the design procedure described in earlier sections, we designed a set of codebooks of size varying from 16 to 4096 by the generalized Lloyd algorithm [ 9 ] with respect to mean squared error distortion measure.”
Small-footprint high-performance deep neural network-based speech recognition using split-VQ (2015, methodology): “The optimal codebook can be found using the LBG algorithm [23], which is summarized in algorithm 1.”
Multi-task learning for sequential data via iHMMs and the nested Dirichlet process (2007, methodology): “The feature vectors across all the ten clips are concatenated to perform vector quantization (VQ) (Linde et al., 1980), mapping each feature vector to a code within a VQ codebook of size 32.”
Recursive Least Squares Dictionary Learning Algorithm (2010, methodology): “, the (generalized) Lloyd algorithm (GLA) which is also known as the LindeBuzo-Gray (LBG) algorithm, update the codebook only after (re)-classification of all the training vectors in the training set [15]–[18].”
Efficient Multi-Start Strategies for Local Search Algorithms (2009, methodology): “In certain cases, gδ(n) = O(e −αn), as shown by Nesterov (2004) for (gradient-based) optimization for convex functions, by Gerencsér and Vágó (2001) for noise-free SPSA for convex functions, or by Kieffer (1982) for k-means clustering (or Lloyd’s algorithm) in one dimension for log-concave densities”
Mathematical Aspects of Divergence Based Vector Quantization Using Fr´ echet-Derivatives (2009, methodology): “for prototype update of the winning prototype ws(v) according to (4.1), also known as the online variant of LBG-algorithm (C‐means) [ 39 ],[65].”