Probabilistic reasoning has become a central part of artificial
intelligence.  
Agents that use probability models can deal effectively with the
uncertainty in the world.
New languages have recently been developed for representing
<a href="prob.html">probabilistic models of complex systems</a>.
A key question is, how can an agent learn such rich probabilistic
models automatically from data?  
With several colleagues, I have begun to investigate this question,
and the results are quite promising:

<ul>
<li>
<a href="../Papers/lprm.ps">Learning Probabilistic Relational Models,</a> N. Friedman, L. Getoor, D. Koller
and A. Pfeffer. <i>Proceedings of the 16th International Joint Conference
on Artificial Intelligence (IJCAI)</i>, Stockholm, Sweden, August 1999.</li>

<li>
<a href="../Papers/ijcai97.ps">Learning probabilities for noisy first-order rules,</a>
D. Koller and A. Pfeffer. <i>Proceedings of the 15th International Joint
Conference on Artificial Intelligence (IJCAI)</i>, Nagoya, Japan, August
1997, pages 1316--1321.</li>
</ul>