#bernoulli-nb

Question 1: Select all that apply

A Bernoulli naive Bayes spam filter uses binary features for $d$ vocabulary words (present or absent). Which statements are true? Select all that apply.

For the two classes it has to estimate $2d + 1$ probabilities1
It should only be used if the words really are conditionally independent given the class2
Naive Bayes can only be used when all features are binary3
Dropping the shared denominator $P(\text{message})$ does not change which class gets the higher score4

Timed test