som: Kohonen's self-organising maps

Description

Self-organising maps for mapping high-dimensional spectra or patterns to 2D; the Euclidean distance is used. Modelled after the SOM function in package class.

Usage

som(data, grid=somgrid(), rlen = 100, alpha = c(0.05, 0.01),
    radius = quantile(nhbrdist, 0.67), init,
    toroidal = FALSE, FineTune = TRUE, keep.data = TRUE)

Arguments

data

a matrix, with each row representing an object.

grid

a grid for the representatives: see 'somgrid'.

rlen

the number of times the complete data set will be presented to the network.

alpha

learning rate, a vector of two numbers indicating the amount of change. Default is to decline linearly from 0.05 to 0.01 over rlen updates.

radius

the initial radius of the neighbourhood to be used for each update: it decreases linearly to 1 over rlen updates. The default is to start with a value that covers 2/3 of all units.

init

the initial representatives, represented as a matrix. If missing, chosen (without replacement) randomly from 'data'.

toroidal

if TRUE, the edges of the map are joined. Note that in a hexagonal toroidal map, the number of rows must be even.

FineTune

apply kmeans for fine-tuning the codebook vectors.

keep.data

save data in return value.

Value

an object of class "kohonen" with components
gridthe grid, an object of class "somgrid".
changesvector of mean average deviations from code vectors.
codesa matrix of code vectors.
classifwinning units for all data objects.
toroidalwhether a toroidal map is used.
datadata matrix, only if keep.data is TRUE.
typethe type of som, here "som"

References

W.J. Melssen, R. Wehrens, and L.M.C. Buydens. Chemom. Intell. Lab. Syst., 83, 99-113 (2006)..

Examples

Run this code

data(wines)
set.seed(7)

training <- sample(length(wine.classes), 120)
Xtraining <- scale(wines[training,])
som.wines <- som(Xtraining, grid = somgrid(5, 5, "hexagonal"))

Xtest <- scale(wines[-training,],
               center = attr(Xtraining, "scaled:center"),
               scale = attr(Xtraining, "scaled:scale"))

som.prediction <-
  predict(som.wines, newdata = Xtest,
          trainX = Xtraining,
          trainY = classvec2classmat(wine.classes[training]))
table(wine.classes[-training], som.prediction$classif)

Run the code above in your browser using DataLab

Get 50% off unlimited learning