------------------------------------------------------
             QwikNet v2.23
       Copyright (C)1996-1999 Craig Jensen
------------------------------------------------------

INTRODUCTION
------------ 
QwikNet is a powerful, yet easy to use artificial 
neural network simulator.  QwikNet implements several 
different algorithms to train a standard feed-
forward neural network in an easy-to-use graphical 
environment.  The shareware version is limited to 
a maximum of 1 hidden layer with 10 hidden neurons
and a maximum of 500 patterns per training data file.  
The registered version can accommodate up to 5 hidden 
layers with no limits (besides hardware) on the 
number of hidden neurons or training patterns.  

You may evaluate this software free for 30 days, 
after which you must either purchase the registered 
version or delete the software from your machine.

Please feel free to distribute the ZIP file 
to friends, associates, or to a computer bulletin board 
system (BBS).


REGISTRATION
------------
There are two ways to register.  You can print and fill 
out the registration form in the file 'register.txt',
(also found in the help file), and mail it with your 
US cash/check payment directly to the author.  Or you 
can register via Kagi Software by running the file 
'register.exe' and following the instructions.  Kagi 
accepts many different payment methods including credit 
cards and non-US funds.  The registration fee is US $49.
Please include $10 US if you want the software on a CD.
Upon receipt of payment you will receive the full 
registered version and all future upgrades for one year.  


INSTALLATION
------------
The 32-bit version, QwikNet.exe, requires a minimum 
486 processor and Windows 95/98 or NT 4.0.  This version removes
all practical limits on network/data file size.  A complete
installation requires less than 5 MB of disk space.

To install QwikNet unzip the file qwiknet.zip and 
run the file setup.exe.


EXAMPLE DATA FILES
------------------
Several example data files are included for testing:

SINCOS.TRN  -  	Training data file with one input (x)
		and two outputs Sin(x) and cos(x).
SINCOS.TST  -   Testing data for SINCOS.
SINCOS.WTS  -	Example of saved weights.
SINCOS.NET  - 	SINCOS project file.

SECURITY.*  -	A more complex data set for Power System Security
		analysis.  The data set contains 33 inputs 
		(power system features) and 1 output 	
		(security classification).

XOR.*       -   XOR training data set.
 		Two inputs and 1 output.

SPIRALS.*   -   The popular "Two spirals" data consisting of 
		data coiled three times around the origin and 
		around one another in the x-y plane.  This is
		a very difficult task for back-propagation 
		networks and their relatives.  

**** NOTE: Acceptable performance on the two spirals problem
           may be impossible given the limitations of the 
           shareware version of QwikNet.  A 3D contour map
           of the solved spirals problem is included in the 
	   QwikNet help file, it uses two hidden layers.


QUESTIONS/COMMENTS
------------------
Please send any questions, comments or bug reports to:

	Craig Jensen
	2488 239 Place NE
	Redmond, WA 98053
	USA
	email - cjensen@kagi.com


