#294: independence
spent a lot of time implementing two classes of probabilistic graphical model today: Markov Networks and Bayesian Networks.
probabilistic graphical models are used to exploit structure in statistical relationships to reduce complexity, facilitate statistical modeling and model-fitting, and perform queries - for example, if you had fitted a PGM on diseases and symptoms, you might present your PGM with evidence (e.g., "I have an upset stomach and my hair turned orange,") and query things from your network (e.g., "What is the single most likely illness, given these symptoms?" to which your PGM might conclude, "You have bad taste in truck stop egg salad sandwiches.")
#294: independence
spent a lot of time implementing two classes of probabilistic graphical model today: Markov Networks and Bayesian Networks.
probabilistic graphical models are used to exploit structure in statistical relationships to reduce complexity, facilitate statistical modeling and model-fitting, and perform queries - for example, if you had fitted a PGM on diseases and symptoms, you might present your PGM with evidence (e.g., "I have an upset stomach and my hair turned orange,") and query things from your network (e.g., "What is the single most likely illness, given these symptoms?" to which your PGM might conclude, "You have bad taste in truck stop egg salad sandwiches.")