In search of sites to host demonstration of distributed belief networks

Robert Dodier (dodier@beethoven.Colorado.EDU)
Thu, 3 Jun 1999 10:31:34 -0600 (MDT)

Hello all,

I am working on a PhD in Civil Engineering here at the University of
Colorado, and my topic is diagnosis and prediction in engineering systems
using distributed belief networks.

I believe it would give my defense a little more oomph if I were able
to demonstrate a working distributed belief network. So I wonder if anyone
out there would like to help me out a little by running on their site the
software which I have cooked up. I've successfully connected belief net-
works on several machines here at CU, but I believe it would be more
convincing to connect b.n.'s at different locations.

In order to get the software running, one need only download and execute
a small stub program, which then pulls the rest of the code across the
Internet. As the system is implemented in Java, this does require install-
ation of the Java interpreter, although that is not difficult.

I would create small belief networks on your site and link together b.n.'s
at different sites to create a single, larger b.n. This doesn't require
file i/o access on your site, but it does require the creation of socket
connections. I believe the security risk is about the same as for running
a web server on your site.

The belief networks I have in mind are very simple diagnostic and
predictive models. The computational burden would be very small, and
network traffic sporadic.

You might wish to write to Michelle Franz (franzm@solnlogic.com), who has
been kind enough to install and run the distributed belief network software
on her site. She could tell you whether it has been burdensome in any way.
She can also vouch for my integrity.

You can find out more about my project at this web site:

http://sonero.colorado.edu/~robert

Please contact me if you are interested. I would be happy to return the
favor, of course.

Have an excellent day,
Robert Dodier