xor.readme

著名的神经网络工具箱

=============================================================================
	README file for the example files xor.xxx
=============================================================================


Description:	This example shows how to train the XOR-problem 
============    on a neural network.


Pattern-Files:	xor.pat
==============	xor-rec1.pat	
		xor-rec2.pat

Xor is the pattern-file for the standard feedforward XOR-network, the
other pattern files are for recurrent networks solving the XOR-problem.


Network-Files:	xor.net
==============	xor-rec.net

xor.net topology: (2-1-1)-feedforward-Network with shortcut
connections from the input units to the output unit.  xor-rec.net is a
recurrent network solving the XOR-problem with one or two time delays
between input and output. It was trained with backpropagation through
time (BPTT). See the user manual in the chapter Neural Network Models
and Functions, section 'Backpropagation through time' for
a more detailed description.


Config-Files:	xor.cfg
=============	xor-rec.cfg

The config files contains the configuration for a 2D display in both
cases.


Hints:
======

The following table shows some learning functions one may use to train
the network. In addition, it shows the learning-parameters and the
number of cycles we needed to train the network successfully. These
parameters have not been obtained with extensive studies of
statistical significance. They are given as hints to start your own
training sessions, but should not be cited as optimal or used in
comparisons of learning procedures or network simulators.

Learning-Function               Learning-Parameters       Cycles

Backpropagation                 2.0                       2000
Backpropagation with momentum   3.0 0.8 0.1               100
Quickprop                       0.3 2.25 0.0001            50
Rprop                           0.2


XOR is a nice small problem to learn about neural networks and
learning algorithms. Try to build a 2-2-1 xor-network and a
2-3-1-network (both without shortcuts) and compare their training
times with the 2-1-1 network with shortcuts given above!


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