Probabilistic Models Quiz
Check your understanding of mixture models and EM — what each step computes, what convergence is (and isn't) guaranteed, and the assumption that makes Naive Bayes naive.
Exercise
How does a GMM's cluster assignment differ from K-Means?
Need a hint?
A point near the boundary between two Gaussians belongs to… both?
Exercise
What does the E-step of EM compute for a GMM?
Need a hint?
Given current parameters, how responsible is each Gaussian for each point?
Exercise
And the M-step?
Need a hint?
It's a weighted version of fitting each Gaussian to 'its' points.
Exercise
What does EM guarantee about the data log-likelihood?
Need a hint?
Guaranteed progress, not a guaranteed destination.
Exercise
What is the 'naive' assumption in Naive Bayes?
Need a hint?
How does it factor P(x₁, x₂, …, xₙ | y)?
Exercise
The EM algorithm fits a GMM by alternating E and M steps. In the ML project loop, EM is playing the role of which slot?
Need a hint?
It's the procedure that searches for the parameters that best fit the data.
Exercise
Transfer test — a technique this course did NOT teach: an attention mechanism replacing a fixed context vector in a sequence model. Which slot does it modify?
Need a hint?
It changes what the model can represent about relationships in the input.