wrapper: wrapper

In mugiro/elm: Extreme Learning Machine for regression and classification

Description

wrapper creates and trains an object of class ELM for given X and Y.

Usage

wrapper(x, y, x_val = NULL, y_val = NULL, neur_type, nn, w = NULL,
  b = NULL, type = "reg", tune = "none", validation = "none",
  ranking = "random", ridge, ....)

Arguments

x	The input data values in a matrix or vector
y	The output data values in a matrix or vector
x_val	The input data values in a matrix or vector for performing a simple validation procedure
y_val	The output data values in a matrix or vector for performing a simple validation procedure
nType	The types of activation functions used in a vector
nNeurons	A vector containing the number of hidden neurons per type of activation function
W	A list of suitable matrix with the input weight vectors or centroids (rbf) per type of activation function
B	A list of suitable vector with the input biases or sigmas (rbf) per type of activation function
structureSelection	A numeric vector with the number of hidden neurons added.
...	Optional additional parameters. None are used at present.

Details

This function is a wrapper for summarizing several actions requiered when creating and adjusting an ELM model. The particular steps are listed below.

1. Creates the SLFN object by calling new().

2. Adds the different hidden neurons by making sequential calls to addNeurons(), one call per each type of activation function defined.

3. Trains the SLFN and obtaines the output weigth vector by calling train().

Value

An object of class "SLFN" with the model developed

Examples

X = seq(0, 7, 0.01)
Y = sin(X)
a = trainELM(X = X, Y = Y, nType = "sigmoid", nNumber = 20)
a = trainELM(X = X, Y = Y, nType = c("sigmoid", "tanH"), nNumber = c(20, 10))

mugiro/elm documentation built on May 23, 2019, 8:21 a.m.