Quote: Original post by InnocuousFox
It's a decent way. Sure, on a small scale, the problem is tractable and you can just brute-force it as you suggested. However, that doesn't scale well.
It's not so much "Bayesian vs. something else" as "how you represent your PDF" (e.g., as a Bayes net)... since representing the whole PDF without any assumptions has the same complexity as brute-forcing (and is in the 0-1 case entirely equivalent).
Quote: Original post by alvaro
If you are fed the results one by one, you can have a container with the surviving possibilities, and filtering from it will get faster and faster as its size is reduced.
...which is just a sparse representation of your Bayesian probability distribution...
Also: Doesn't Wikipedia give an algorithm?
