nadja in wbo

Belief propagation is one of feasible algorithms in order to assess MPM estimation for an acyclic graphical model.

24042008.pdf

This presentation is originally that I talked in the informal seminar which is called "次世代アルゴリズム勉強会" at 05/07/2006. (Note that the seminar site of "次世代アルゴリズム勉強会" is missed now.) This theme is, in probabilistic inference, one of feasible algorithms handling the marginal probability, the estimation is so-called maximum posterior marginal (MPM), and the assessment of covariance with respect to acyclic graphical models. Therefore this scheme is called the other names in some research fields. For instance,

1. Message passing algorithm
   
2. Max product algorithm
   
3. Sum product algorithm
   
4. Survey propagation
   
5. Viterbi decoding algorithm
   
6. Transfer matrix method, etc.
   

Have you ever known them?