Finding the current state from HMM

RyuParticipant January 16, 2012 at 5:00 pm
In the SuanShu example, MarkovModel.java, I found the following code is to train the parameters A and B
```
for (int i = 1; i < = nIterations; ++i) {
       model_0 = new HmmTrainByEM().train(model_0, observations);
}
```
How can I find the best estimated states from the trained model, such like by Viterbi algorithm, in algoquant? That means if there are 100 observations to train the matrices A and B, is there a method in model_0 to provide the best fit states for those 100 observations? Thanks a lot.
RyuParticipant January 17, 2012 at 3:04 am

Hello,

We currently do not have the Viterbi algorithm in SuanShu. It would be useful to have it in our math collection. We add this to our TODOs.

Thanks.
RyuParticipant January 19, 2012 at 2:42 am

I see. Just wonder whether there would be other methods to find the states in HMM. Thanks.