|
|
|
 |
 |
 |
 |
 |
 |
Plenary Speakers 
|
 Stuart Geman Probabilistic Grammars and their Applications I will review the theory of formal grammars. The simplest grammars are regular grammars, which become Markov chains when fitted with a probability distribution. More general and more powerful are the context-free grammars, the probabilistic versions of which are equivalent to branching processes. Most applications in language, vision, and neural modeling appear to require even more general grammars: context-sensitive grammars or grammars of still higher order. There is no general theory for how to put probability distributions on context-sensitive grammars, yet it is increasingly apparent that successful applications require a probabilistic framework, with its associated statistical theory and theory of inference. I will suggest a method for placing workable probability distributions on context-sensitive (and more general) grammars, and I will discuss some applications.
Stuart Geman received his Ph.D. in Applied Mathematics from MIT in 1977. He joined the faculty at Brown in 1977, where he has been Professor of Applied Mathematics since 1985, and the James Manning Professor since 1997. He received the Presidential Young Investigator Award in 1984 and is a Fellow of the IMS. Professor Geman's research has been in nonparametric inference, the theory and applications of Markov random fields, and Bayesian methods for image analysis and vision. He has worked extensively with industrial applications of computational vision. His current interests are in compositional models for vision, coding theory, statistical methods for analyzing multi-unit neuronal recordings, and theories of neural representation. More Information Contact conald@cs.cmu.edu for more information The conference is sponsored by CMU's newly created Center for Automated Learning and Discovery. |
