CLSP Guest Lecture Series
|[ WS99 Homepage ]|
AT&T Labs Research
Artificial intelligence (AI) has come a long way in the last decade, and is now a field with firm mathematical foundations and increasing ambitions towards difficult, large-scale applications. Two important trends are the widespread adoption of mathematical models with explicit representations of uncertainty and stochastic behavior, and the development of models that can succinctly exploit natural structure in very large or complex domains. The move towards rich probabilistic models has been accompanied by the discovery of a number of elegant algorithms for extracting the knowledge these models contain, and for learning the models from observed experience.
In this talk, I will survey these developments as they have been investigated in Markov decision processes and Bayesian networks. This line of research highlights the connections between "modern" AI and many older fields, such as statistics, control theory, statistical physics, and economics.