MANFRED JAEGER and EMIL WEYDERT Max-Planck-Institut fuer Informatik, Saarbruecken The course explores various frameworks for probabilistic reasoning
in formal systems, and their application in artificial intelligence.
The study of systems that combine the representational rigor of
formal logic with the processing of quantitative information goes
back as far as Rudolf Carnap's work on inductive inference. In
the context of artificial intelligence there has been a renewed
and increasing interest in such formalisms. We will discuss various logical systems for reasoning about probabilities,
ranging from simple systems for processing linear constraints
on the probabilities of propositional formulas, to very expressive
extensions of first-order and modal logics. We will discuss the nonmonotonic aspect of commonsense probabilistic
reasoning, its similarities with nonmonotonic deductive reasoning,
and its implementation by entropy maximization strategies. The course will also provide an introduction to Bayesian networks
as a means of representation that often makes the inference problem
tractable.
FORMAL SYSTEMS FOR PROBABILISTIC INFERENCE
jaeger@mpi-sb-mpg.de and weydert@mpi-sb-mpg.de
None
Textbooks on the topic of this course are fairly scarce. "The
Uncertain Reasoner's Companion" by J.B.Paris contains a part of
the material we intend to present, and approximateley indicates
the level of the course.