=============================================================================
	README file for the example files spirals*.xxx
=============================================================================


Description: This network is an example of using the RBF DDA algorithm
============

The task was to solve the two spirals porblem.


Pattern-Files:	spirals.pat
==============	

The pattern   file defines the  well known  spiral problem.  The input
consists   of two  values (x  y   position) which range  from  -6.5 to
6.5.  The output  contains two  values which  classify the two spirals
(values 0 and 1).


Network-Files:	spirals_dda.net
==============	

The  network  contains  a   trained network file  with   the following
topology:
	2  input neurons (organized as 3x3 input mask)
	45 RBF hidden neurons
	2  output neuron


Config-Files:	spirals_dda.cfg
=============

This network uses one 2D display in its standard configuration.


Hints:
======

The network is already trained  by the RBF_DDA learning function. Note
that the RBF_DDA algorithm assumes a winner takes all behaviour in the
output layer. Therefore the output (3.4 1.2) of the two output neurons
is a correct output for the training pattern (1  0). Creating a result
file or  using other learning algorithms  will result in strange error
values.

To retrain  this network all hidden neurons  must be deleted  by using
the  graphical  network editor (select all   hidden units and give the
command 'Units Delete'). A  valid set of  learning parameters would be
0.4 0.2 5,  which defines two threshold  values and  the size of  unit
columns in the new created hidden layer.

Pressing ALL several times creates enough RBF hidden neurons two learn
the spirals task with 100% accuracy.

For more  information about the RBF DDA  algorithm please refer to the
user manual or to the author berthold@ira.uka.de.


=============================================================================
	End of README file
=============================================================================