muHVT (version 3.0.0)

VQ_codebookSplit: VQ_codebookSplit

Description

Vector Quantization by codebook split method

Usage

VQ_codebookSplit(dataset, quant.err = 0.5, epsilon = NULL)

Value

clusters

List. A list showing each ID assigned to a cluster.

nodes.clust

List. A list corresponding to nodes' details.

idnodes

List. A list of ID and segments similar to nodes.clust with additional columns for nodes ID.

error.quant

List. A list of quantization error for all levels and nodes.

plt.clust

List. A list of logical values indicating if the quantization error was met.

summary

Summary. Output table with summary.

Arguments

dataset

Matrix. A matrix of multivariate data. Each row corresponds to an observation, and each column corresponds to a variable. Missing values are not accepted.

quant.err

Numeric. The quantization error for the algorithm.

epsilon

Numeric. The value to offset the codebooks during the codebook split. Default is NULL, in which case the value is set to quant.err parameter.

Author

Sangeet Moy Das <[e-mail]>

Details

Performs Vector Quantization by codebook split method. Initially, the entire dataset is considered to be one cluster where the codebook is the mean of the cluster. The quantization criteria is checked and the codebook is split such that the new codebooks are (codebook+epsilon) and (codebook-epsilon). The observations are reassigned to these new codebooks based on the nearest neighbour condition and the means recomputed for the new clusters. This is done iteratively until all the clusters meet the quantization criteria.

See Also

hvtHmap

Examples

data("iris",package="datasets")
iris <- iris[,1:2]

vqOutput = VQ_codebookSplit(iris, quant.err = 0.5)