Elm.cross.valid: Training ELM with cross validation.

In ForAI: Forecasting with Artificial Intelligence

Description

A k-fold cross validation on the training data is performed.

Usage

Elm.cross.valid(X.fit, Y.fit, Number.hn=10, n.blocks=5, returnmodels = FALSE, 
                            autorangeweight=FALSE, rangeweight=1, 
                            activation='TANH',outputBias = FALSE,
                            rangebias = 1)

Arguments

X.fit	Data matrix (numeric) containing the input values (predictors) used to train the model.
Y.fit	Response vector (numeric) used to train the model.
Number.hn	Number of units in the hidden layer. Default is 10.
n.blocks	an integer specifying the desired number of cross-validation folds. Default is 5.
returnmodels	whether to return the trained models. Default is FALSE.
autorangeweight	Option whether to use the automated range used for the weights. Default is FALSE.
rangeweight	Initial random weights on [-rangeweight,rangeweight]. The default is 1

activation	Activation function of the hidden layer neurons. Available functions are: 'TANH' (default) and 'SIG'.
outputBias	Option whether to use the bias parameter in the output layer
rangebias	Initial random bias on [-rangebias,rangebias]. The default is 1

Value

If returnmodels=FALSE (default), it returns the fitted values. If returnmodels=TRUE, it returns a list containing:.

predictionTrain	The fitted values for the training data.
trained.elms	A list containing the k models used.

Examples

set.seed(123)
attach(wtloss)
library("scales")

#scaling the inputs/outputs
x.train <- rescale(as.matrix(wtloss$Days), to=c(-1,1))
y.train <- rescale(as.matrix(wtloss$Weight), to=c(-1,1))

#training the ELM
elm.fit.values <- rescale(Elm.cross.valid(x.train,y.train,Number.hn=5,n.blocks=5),
                          to= range(as.matrix(wtloss$Weight)),from=c(-1,1))

oldpar <- par(mar = c(5.1, 4.1, 4.1, 4.1))
plot(wtloss$Days, wtloss$Weight, type = "p", ylab = "Weight (kg)",
      main="Weight Reduction",pch=20)
points(wtloss$Days, elm.fit.values,col=2,type='p',pch=20)

ForAI documentation built on May 2, 2019, 6:14 p.m.