Oct 15
On Tuesday I assigned two problems to consider :
- a variation on AIMA 13.13 ("what happens if we both tests?")
- spam filtering, 13.22
We'll work through both in class.
My solution to the "sick" problem is attached. Note particularly
the distinctions between between "conditionally independent"
and "independent" for the two tests, and that
P(A|S,B) = P(A|S) # A depends on S, not (directly) on B
P(A|B) != P(A) # but A does depend (indirectly) on B
Then we'll talk about naive bayes spam filtering. See
The basic idea is a straightforward generalization
of the Bayes approach, however there are several
tricky bits, namely (a) how to combine the probabililties
and (b) what to do about low probability words
(P(word)=0 is problematic).
See you in a week, when we'll continue
working our way through chapters 13,14,18 in the text
and other sources.
To think about :
Do the spam problem with two words, and
see if you can get the new formulas
to make sense with what we did
for the "sick" problem with two tests.
![[paper clip]](/courses/source/wiki_images/paper_clip_tilt.png)
name last modified size