Probability theory provides a sound basis for reasoning under
uncertainty.
We can represent an agent's uncertainty about the world using a
probability model,
and the agent can use the model to draw likely conclusions from its
observations.
Daphne Koller and I, together with several other colleagues, have
developed
a powerful new language for representing probability
models of rich, complex systems.
We have also developed efficient algorithms for reasoning with
these models.
Our language is based on Bayesian networks, but is much more expressive.
It allows the model designer to describe a complex system in terms of
objects and the relationships between them, and to describe how
properties of an object depend probabilistically on properties
of related objects.
The use of a structured representation language makes the
representation of complex probability models much easier.
As it turns out, it also makes probabilistic inference much more effective.
Selected papers:
-
SPOOK: A System for Probabilistic Object-Oriented Knowledge Representation,
A. Pfeffer, D. Koller, B. Milch and K.T. Takusagawa.
Proceedings of the 14th Annual Conference on Uncertainty in AI
(UAI),
Stockholm, Sweden, July 1999.
-
Object-Oriented Bayesian Networks, D. Koller
and A. Pfeffer.
Proceedings of the 13th Annual Conference on Uncertainty
in AI (UAI), Providence, Rhode Island, August 1997, pages 302--313.
-
Probabilistic frame-based systems, D. Koller
and A. Pfeffer. Proceedings of the 15th National Conference on Artificial
Intelligence (AAAI), Madison, Wisconsin, July 1998.
-
Effective Bayesian Inference for Stochastic Programs,
D. Koller, D. McAllester, and A. Pfeffer.
Proceedings of the 14th National
Conference on Artificial Intelligence (AAAI), Providence, Rhode Island,
August 1997, pages 740--747.
You can also get the complete scoop on this topic from my PhD thesis: