FFNC

Feed-forward neural net classifier back-end

    [W,HIST,UNITS] = FFNC(ALG,A,UNITS,ITER,W_INI,T,FID)

Input
ALG	Training algorithm: 'bpxnc' for back-propagation, 'lmnc'  for Levenberg-Marquardt
A	Training dataset
UNITS	Array indicating number of units in each hidden layer.  Default is a single hidden layer. Its size is the half of the  number of objects in A divided by feature size plus class size  (roughly half of the number of parameters to be optimised) with  a maximum of 100;
ITER	Number of iterations to train (default: inf)
W_INI	Weight initialisation network mapping (default: [], meaning  initialisation by Matlab's neural network toolbox)
T	Tuning set (default: [], meaning use A)
FID	File ID to write progress to (default [], see PRPROGRESS)

Output
W	Trained feed-forward neural network mapping
HIST	Progress report (see below)

Description

This function should not be called directly, but through one of its  front-ends, BPXNC or LMNC. Uses the Mathworks' Neural Network toolbox.

This routine escapes to KNNC if any class has less than 3 objects.

See also

mappings, datasets, bpxnc, lmnc, neurc, rnnc, rbnc, knnc,