CN710: Bayesian Belief Networks

CN710: Bayesian Belief Networks


Chaitanya Sai Gaddam

Eugene Zaydens

03/27/20006



Bayesian belief networks are one class of probabilisti
c graphical models that has been
popular among machine learning researchers with varied interests. They allow for an
i
ntuitive

understanding of dependencies among factors
,

and Bayesian inferences make
the construction of joint
probabilities of all observable factors tractable. Like decision
trees, the process resulting in a certain classification is transparent and the le
arned
structures can aid expert diagnosis.
The
articles we have chosen for the discussion cover
the fundamental concepts behind belief networks and briefly touch upon the
mathematical details
(and computational problems)
of
using these networks for infere
nce
and in learning the structure of dependencies. The antiterrorism article illustrates the use
of Bayesian networks in a domain where understanding of the classification technique is
as important as the classification performance.


Core Readings:


Char
niak
, E

(1991)

Baysian networks without

tears.

AI Magazine
, 12(4), Winter
.



Niedermayer
, D

(1998)
An introduction to Bayesian Networks and
their
Contemporary
Applications
[WWW document]

URL
http://www.niedermayer.ca/papers/bayesian/


Hudson, L. D., Ware,

B. S., Laskey, K. B. & Mahoney, S. M. (unpub.),
An Application
of
Bayesian Networks to Antiterrorism Risk Management for Military Planners
.


Supplementary Readings:


Heckerman, D (1994)

A Tutorial on Learning with Bayesian Networks
,
Technical

Report
MSR
-
TR
-
95
-
0
6
,
Microsoft

Research


We very strongly urge you to read this
article

despite
its acid
-
reflux inducing
math density



Kahney, L (2001)
MS Office Helper Not Dead Yet

[WWW document]

URL
http://www.wired.com/news/technology/0,43065
-
0.html

Read this article whenever you are overwhelmed by the urge to tear up the
aforementioned article


Krieg, L. M (2001) A Tutorial on Bayesian Belief Networks,
Technical Report DSTO
-
TN
-
0403
, Electronics

and Surveillance Research Laboratory


A huge and comprehensive tutorial that also explains basic probability (60 pages)