Yeah, and it uses the same analogy for understanding belief propagation as Pearl himself uses, and a reference to Pearl, and a bit more discussion of Bayes nets as a good way to understand things. But, I think, a lot of people didn’t derive the directive “Learn Bayes nets!” from that example of insight derived from Bayes nets (and would benefit from going and doing that).
I do think there are some other intuitions lurking in Bayes net algorithms which could benefit from a similar write-up to Fake Causality, but which went “all the way” in terms of describing Bayes nets, rather than partially summarizing.
Fake Causality contains an intuitive explanation of double-counting of evidence.
Yeah, and it uses the same analogy for understanding belief propagation as Pearl himself uses, and a reference to Pearl, and a bit more discussion of Bayes nets as a good way to understand things. But, I think, a lot of people didn’t derive the directive “Learn Bayes nets!” from that example of insight derived from Bayes nets (and would benefit from going and doing that).
I do think there are some other intuitions lurking in Bayes net algorithms which could benefit from a similar write-up to Fake Causality, but which went “all the way” in terms of describing Bayes nets, rather than partially summarizing.