Met.NN2 {Metabonomic}R Documentation

Neural Network 2

Description

Creates a feedforward artificial neural network according to the structure established by the AMORE package standard.

Usage

Met.NN2(datos, externa)

Arguments

datos Spectra data frame
externa Not implemented yet

Details

This application creates a feedforward artificial neural network according to the structure established by the ''AMORE'' package. With this application the user can select the number of layers and the number of neurons in each layer. Moreover, a lot of parameters can be controlled by the user like the learning rate at which every neuron is trained, the momentum for every neuron, the error criterion (Least Mean Square or Least Mean Logarithm Squares), the activation function of the hidden and the output layer (Purelin, Tansig, Sigmoide or Hardlim) and the training method (Adaptative gradient descend or BATCH gradient descend, with or without momentum). With these parameters selected, the algorithm trains the network with the samples manually or randomly selected, and it tests the networks with the other samples. Launched with the GUI. Beta version.

Author(s)

Jose L. Izquierdo izquierdo@ieb.ucm.es

References

AMORE package http://finzi.psych.upenn.edu/R/library/AMORE/html/newff.html


[Package Metabonomic version 3.1.2 Index]