🚀 go-pugleaf

ALN Educational Kit

If you are interested in learning about adaptive logic networks
(ALNs), but the problem is you don't have up-to date software (Atree
2.0 and Atree 2.7 are fairly obsolete now), then the Atree 3.0
Educational Kit is just what you need.  You can ftp it from
ftp.cs.ualberta.ca in pub/atree/atree3/atree3ek.exe (binary mode,
approx 1.2 Mb).

The software runs under Windows (3.1, NT Server, NT Workstation, 95,...).

One remark about the setup file you get: it may ask for some responses
when it finds some software components in use.  Just click on "ignore"
in that case.

The ALN Educational Kit handles functions of one or two input
variables (where the Atree 3.0 beta package is not restricted in this
way.)  This means that you can view and study the result of learning
using the included gnuplot software.  The two-spirals problem is one
of the examples, but there are other examples that illustrate how you
can include your knowledge of the *qualitative* properties of a
function to limit the admissible solutions.  For example, you can
force functions to be convex-up (dome-shaped) by using a very simple
net architecture.

For projects, the kit includes software that can initiate ALN learning
or evaluation from any program that can call a DLL.  Examples show
how this can be done from MS-Excel and from MS-Access, with an example
where you try to predict the opponent's bid in a competitive bidding
situation.

Time shifts are included as a feature of the ALN Definition Language
(ADL), for those who deal with time-series or control problems (and of
course those who play the stock market).

Atree 3.0 has been used very successfully to give real-time functional
electrical stimulation to control a walking prosthesis for incomplete
quadriplegics.  Other promising areas of application where the student of
ALNs can do some exciting projects are in predictive maintenance,
nondestructive testing, pattern recognition, computer vision.

Most real problems will require the full Atree 3.0, and, without
becoming commercial here, the good news is that it costs less than $100
US to acquire while it it still under beta test (probably until October '95).
You can email me whenever you decide to upgrade (arms@cs.ualberta.ca)
and I'll provide details.

We would like to keep improving the Educational Kit, and your suggestions
are welcome.

Instructors of neural net courses who would like to include an ALN
component can inquire about special rates for upgrades for students
doing class projects.

Have fun with ALNs using the Educational Kit!  In a subsequent posting,
I'll put out a one-pager that shows how to set up a net and run tests.

Bill

--
* 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