Forward-Backward log_n_states

[Lucas0T]

Hi,

I'm wondering why in forward-backward and viterbi you divide by the number of states
log_n_states?
As far as I can see it is not part of the standard viterbi or forward-backward algorithm.
Or am I overlooking something?
The lines are:
Forward_Backward.hpp line 64:
cell(i, j).alpha = pm.log_pr_corrected_emission(j, ev[0]) - log_n_states;
Viterbi.hpp line 60:
cell(0, j).alpha = pm.log_pr_corrected_emission(j, ev[0]) - log_n_states;

Best Regards