hmmdecode performs forward-algorithm?

Hi, my question is basically if the command hmmdecode performs the forward-algorithm? I believe the FORWARD output should give me this. But, if I want the output to be the forward-variable for each consecutive time-step, what should I look at?

Related to this, a possible problem is that hmmdecode always starts in the same state, s_1. I have seen some answers to how to get around this, but I found them a bit cumbersome. Is there another way to do this (e.g. other algorithm)?

Answer 1

[Abhishek Krishna]

Hi,

The `hmmdecode` function in the Hidden Markov Model (HMM) toolbox performs the forward algorithm. The forward algorithm calculates the probability of being in a particular state at each time-step, given the observed sequence.

To obtain the forward-variable for each consecutive time-step, you can look at the `Alpha` output of the `hmmdecode` function. The `Alpha` matrix contains the forward probabilities for each state at each time step.

Regarding the issue of `hmmdecode` always starting in the same state, you are correct that it typically assumes the initial state to be fixed (often denoted as s_1). If you want to start from a different initial state, you can modify the initial state probabilities in the HMM model.

I hope this helps!

[Benjamin Andersson]

Hi, do you mean the "FORWARD" output? I don´t think the documentation (https://se.mathworks.com/help/stats/hmmdecode.html) has an output called "Alpha". I was unsure whether PSTATES or "FORWARD" gives what I want, but I now believe it is "FORWARD".