Learn R Programming

som.nn (version 1.4.4)

Topological k-NN Classifier Based on Self-Organising Maps

Description

A topological version of k-NN: An abstract model is build as 2-dimensional self-organising map. Samples of unknown class are predicted by mapping them on the SOM and analysing class membership of neurons in the neighbourhood.

Copy Link

Version

Install

install.packages('som.nn')

Monthly Downloads

158

Version

1.4.4

License

GPL-3

Maintainer

Andreas Dominik

Last Published

April 3rd, 2024

Functions in som.nn (1.4.4)

som.nn.round.votes

Rounds a dataframe with vectors of votes for SOMnn
plot_predictions

Plots predicted samples as points into a plotted som.
som.nn.train

Hexagonal som training
som.nn.validate

Predict class labels for a validation dataset
som.nn.all.accuracy

Calculate overall accuracy
som.nn.confusion

Calculate confusion matrix
som.nn.run.kernel

calls the specified kernel for som training.
som.nn.visual

Mapping function for SOMnn
som.nn.export.som

Export a som.nn model as object of type SOM
som.nn.visual.one

Maps one vector to the SOM
round.probabilities

Advanced rounding of vectors
som.nn.set

Set parameters for k-NN-like classifier in som.nn model
som.nn.multitrain

Multi-step hexagonal som training
som.nn.som.experimental

Work hourse for som training.
som.nn.max.row

Special version of maximum finder for SOMnn
som.nn.continue

Continue hexagonal som training
som.nn.som.gaussian

Gaussian kernel for som training.
som.nn.do.train

Work hourse for hexagonal som training
som.nn.som.internal

Work hourse for som training.
get.border.neurons

Get border neurons.
initialize,SOMnn-method

Constructor of SOMnn Class
dist.fun.inverse

Inverse exponential distance functions for topological k-NN classifier
dist.torus

Torus distance matrix
dist.fun.tricubic

Tricubic distance functions for topological k-NN classifier
dist.fun.linear

Linear distance functions for topological k-NN classifier
dist.fun.bubble

Bubble distance functions for topological k-NN classifier
SOMnn-class

An S4 class to hold a model for the topological classifier som.nn
norm.linear

Linear normalisation
som.nn.accuracy

Calculate accuracy measures
predict,SOMnn-method

predict method for S4 class SOMnn
plot,SOMnn,ANY-method

Plot method for S4 class SOMnn
som.nn-package

Topological k-NN Classifier Based on Self-Organising Maps
norm.softmax

Softmax normalisation
enrich.dirty

enrich training set with dirty mapped samples
hexbinpie

Plots the hexagonals and pi charts. Adapted code from package somplot.
make.codes.grid

Makes a data.frame with codes coordinates
som.nn.export.kohonen

Export a som.nn model as object of type kohonen
makehexbinplot

makes the actual heagonal plot. Adapted code from package somplot.