home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Frozen Fish 1: Amiga
/
FrozenFish-Apr94.iso
/
bbs
/
alib
/
d5xx
/
d510
/
synworks.lha
/
SYNWORKS
/
ReadMe
< prev
next >
Wrap
Text File
|
1991-06-29
|
5KB
|
167 lines
SYNWORKS
A neural network development tool for the CBM Amiga.
Written & Copyright by
Michael Kaiser
Im kleinen Feld 5
3530 Warburg 1
Tel. 05641/4521
or
Schindweg 3
7500 Karlsruhe 41
Tel. 0721/407968
-----------------------------------------------------
SynWorks was developed in addition of an M.S. work in cooperation with
HP, Germany. It is one of the first Amiga tools for "AI-applications",
so it may give the Amiga-market a "kick" in the direction of science.
The PD-Version does not contain all Features of the professional one,
but I think it is very good to get a first impression of what neural
network are and how they could be used.
If you'd like to get some information of neural network theory,
you may read one of the following books/articles:
1. R. Beale, T. Jackson
Neural computing - an introduction
Adam Hilger Bristol, Philadelphia, New York
1. edition 1990
2. Robert Hecht-Nielsen
Neurocomputing
Addison-Wesley,
1. edition 1990
------------------------------------------------------------------------
If you like SynWorks and would be glad to get a full version,
including linkable object files for nn-use in own applications,
manuals and a neural-network tutorial,
please send $70 (US) or DM 120,- (europe) to
Michael Kaiser
Im kleinen Feld 5
D-3530 Warburg 1
West Germany
For additional $30 / DM 50,- you'll also get SynTools incl. manual,
a tool to display and preprocess synworks training & test files.
----------------------------------------------------------------------
INSTALLATION
You need at least 1 MByte of memory.
- NO HARDDISK:
Copy your original CBM WorkBench 1.3 or 2.0 to another disk.
Remove the following:
FONTS: all exept topaz
DEVS: all printers exept the one you possess.
SYSTEM:your choice
TOOLS: your choice
C: your choice,
so that you get about 400 KByte of free disk space.
Modify your startup-sequence to get a RAM-DISK,
reboot your computer.
Insert the SynWorks/Fish-Disk, open it and copy the SYNWORKS-DRAWER
a) to the RAM DISK (if you have one dd)
b) to the prepared WB-Disk.
If you have one disk, re-insert the WB-Disk and copy the SYNWORKS-DRAWER
to it.
- HARDDISK:
Insert the SynWorks/Fish-Disk, open it and copy the SYNWORKS-DRAWER
to the partition you want.
----------------------------------------------------------------------
SETTING UP:
If you start SynWorks the first time, it will probably display an
error message "SynWorks-Title not found" and/or "SynWorks-Screendisplay"
not found.
To complete the installation, ignore these messages and select
"Setup/Preferences" after starting SynWorks.
Set the Path "Synworks-Path" to the one you installed SynWorks in,
also set the Editor "Editor" to the one you want to use.
Select "Use prefs" to save preferences.
Select "Setup/Screendisplay" and modify the given values if you want,
then select "OK" to save.
-----------------------------------------------------------------------
FIRST STEPS:
This text is no manual, but after you have correctly installed SynWorks,
you should play a little bit with its features.
To do this, select "NetWork/Load NetWork", enter the "networks"-directory
(relative to your SynWorks-Path) and doubleclick on the "xor_bp.nwk"
entry. A backpropagation-network soving the XOR-Problem will appear,
which may be used as playground.
----------------------------------------------------------------------
FILE FORMAT:
SynWorks uses ASCII-Files for training and test, normally reading
one input vector from one line in the file.
It knows about some special characters:
"!" : All following chars in the line are comments.
";" : The vector contains also the following line
Look at the XOR.INP file to check this.
---------------------------------------------------------------------------
CONSTRAINS:
The SynWorks-PD Version supports following
network models:
Backpropagation
Perceptron
Kohonen
The number of neurons per layer must be smaller than 11.
--------------------------------------------------------------------------
Last, but not least: Thanks to Fred Fish for all his work !
--------------------------------------------------------------------------