|
Calculates the cross-entropy between the two selected HMM models based on observation sequences.
The cross-entropy is a measure of the distance between two models λ1 and λ2. It is defined as
D(λ1,λ2) = 1/N (log p(O2|λ1) - log p(O2|λ2)), |
where O2 is an observation sequence of length N generated by model λ2.
The symmetrized version is:
Ds(λ1,λ2) = (D(λ1,λ2) + D(λ2,λ1))/2. |
© djmw 20101017