Intro
Admin remember to come by office hours once No class next Tuesday New Material Try and put this all in perspective Slides Our solutions, Forward backward Viterbi EM We're efficient solutions given a particular structure of a graphical model Let's look at a more general structure for the graphical model Bayesian Network Not time dependent Observed variables Unobserved variables Absence of links is where you find efficiencies Links themselves have probabilities associated with them. Bring it back to the case of the Markov Models Bring it back to the case of the Hidden Markov Models Introduce the idea of parameter tying Introduce Dynamic Bayesian Network Create a model of movement through a building that is a DBN