🚀 go-pugleaf

Many thanks to the people who have ftp-ed the Atree 3.0 Educational
Kit (EK) to learn about adaptive logic networks.  This is fabulous
encouragement for us developers.  It would be good guidance for us if
everyone who tries EK would provide some feedback -- either via email
or posting to a newsgroup -- whichever is appropriate.  We would like
to make this Educational Kit as bug-free as possible so students doing
projects will not have unnecessary hassles.

We are continually adding features and fixing bugs, so please ftp a
new version from time to time -- without waiting for an "expired" nag
message.

Have you found any totally new things you can do with ALN neural nets
by including qualitative properties of the functions in setting up
training?  People following the use of a priori knowledge in the
financial area, eg the use of "hints" about monotonicity (as in the
work of Abu-Moustafa) will have a lot to experiment with using ALNs.
Let us hear about it!

The latest bug in EK is that the delete key doesn't work in the editor
(which means you have to use right arrow and backspace to delete).
That should be fixed soon, and better facilities will be added for
loading decisions trees, the mechanism for high speed evaluation of
ALNs.

Bill

PS for those who still haven't tried ALNs, you can get this program
that is the same as the commercial Atree 3.0 software, except for a
limitation to three variables.  EK is not cheap demo advertising, but
software that can do non-trivial problems like the two spirals problem
which are hard for backprop nets (the code is in a sample!!).

EK is to be found on

ftp.cs.ualberta.ca

in

pub/atree/atree3/atree3ek.exe

(~900K, binary mode, for Windows 3.x and higher).
--
* Prof. William W. Armstrong,		arms@cs.ualberta.ca
* Computing Science Department,		tel(403)492 2374 fax (403)492 1071
* University of Alberta,		home (403) 438 1103
* Edmonton, Alberta, Canada T6G 2H1	http://web.cs.ualberta.ca/~arms