Week 4 Flashcards
What are the three main classes of Kalman models?
What is the idea of exponential family models? What is the main result of this idea?
What is the stacked form notation? What do all the components mean?
What are the six different densities? What are their definitions?
What is the difference between the difference between the conditional
observation density and the observation density?
What is available for (non)-linear Gaussian models but not for non-linear non-Gaussian models?
What are the two things that need to be estimated using numerical optimization for non-linear non-Gaussian models?
Why can the mean be used to estimate the relevant parameters in a Gaussian model?
What is the Linear Gaussian Model (in stacked notation)?
What is the analytic expression for the mode in a Linear Gaussian model? Why is it not used in practice?
Which method is used for numerical mode estimation? How does it work?
How is the Bayes rule applied to ensure that the unkown p(theta | Y) can be computed?
What are the two analytic expressions for the Linear and Non-Linear models? What is important about their differences?
What comes after this part?
Give a quick summary of what we do to do mode estimation.
Where is importance sampling needed? (i.e., which types of models)
What is the definition of the conditional expectation? Why can’t this be easily estimated (for non-Gaussian non-linear models)?
How is the importance density generally chosen? What is the main idea of importance weights?
