Probabilistic Graphical Models: Applications in Biomedicine
4Probabilistic graphical models include a variety of techniques based on probability and decision theory-techniques that give us a theoretically well-founded basis for making decisions under conditions of uncertainty and to solve complex problems efficiently. Over the last year, these methods have been used in a great variety of applications, from medical expert systems to intelligent user interfaces. In this talk, I will give a general introduction to probabilistic graphical models and describe some of the most popular ones, such as Bayesian networks and Markov decision processes. Then I will demonstrate their application in three complex problems in biomedicine: (1) helping a physician guide an endoscope in the colon, (2) modeling the evolutionary networks of HIV, and (3) adapting a stroke rehabilitation system for the patient.