Anonymous - 2011-07-12

Hi all,

Does anyone know how to learn some of the parameters in a DBN and freeze the rest of them - In case if I have those parameters a priori?

For example in the HMM learning problem:

X0 -> X1
  |          |
  v        v
Y0     Y1

, what if I already have the observation probability distribution P( Y | X ) and just need to learn the state transition probability function P ( X' | X ) from the evidences?