learn: Train the network (Online algorithm).

In harrysouthworth/kohonen: Self-organizing Maps for Data Classification

Description

Perform the learning process of the network with online algorithm.

Usage

learn( object, number.iter, order, max.alpha = 0.5, min.alpha = 0.0, max.rayon = 3, min.rayon = 0, step.eval.si = 100 )

Arguments

object	a som object
number.iter	number of learning process' step. It won't have effect if order is specified.
order	vector of length number.iter, specifying which row is presented to the learning process at the n-ith step.
max.alpha	the maximum amount of change during learning process.
min.alpha	the minimum amount of change during learning process.
max.rayon	the maximum radius during learning process.
min.rayon	the minimum radii during learning process.
step.eval.si	integer. A vector will be created, it will have values of evenly spaced numbers from 1 to number.iter with spacing step.eval.si. At each value of the vector, an evaluation of energy function will be performed.

Details

Be carefull with step.eval.si, the smallest is the value, the longer it will take time to process. alpha decrease linearly from max.alpha to min.alpha over number.iter updates. rayon decrease linearly from max.rayon to min.rayon over number.iter updates.

Value

a som object.

Author(s)

David Gohel

References

Kohonen, T. (1995). Self-Organizing Maps

See Also

som learnBatch plot.som summary.som predict.som biplot.som getWeights setWeights

Examples

library(MASS)
lcrabs <- log(crabs[, 4:8])

lcrabs.som <- som ( formula = ~ . , data = lcrabs
	, neighborhood = "uniform"
	, grid = grid ( xdim = 10 , ydim = 10 , type = "hexagonal" ) 
	, weights.min = min (lcrabs), weights.max = max (lcrabs)
	)
# train the network
lcrabs.som <- learn( lcrabs.som , number.iter = 500 , max.alpha = 0.3, step.eval.si = 50)
# plot energy function and print a summary
plot( lcrabs.som , "energy")
summary( lcrabs.som )

harrysouthworth/kohonen documentation built on May 17, 2019, 3:03 p.m.