melmasri/traveltimeHMM

Question on forwardbackward

Closed this issue · 1 comments

@ericgermaining in regard to your question here

# QUESTION: 'bwd' does NOT make use of 'fv' from 'fwd' unlike in the paper. The interaction between 'fv' and 'b'

I realize that this is not the most intuitive forward backward implementation, one because probTran is outside the scope of forwardbackward. I do that to not recalculate probTran when creating the probJointStates in here

probJointStates = normalizeR(alpha[-nObs, rep(1:nQ, each = nQ)] *
.

One possible way to make it more intuitive, maybe to create a function like est_state_prob in this function we calculate probTran, pass it to forward and return back probStates and probJointStates? what do you think?

Also I want to confirm did you find any methodological issue in this implementation, other than the scope of forwardback?

@melmasri As an answer to your question: I did not find any further problem. I guess we can close the issue. I transferred your code to the review-stuff branch.