Search All of the Math Forum:

Views expressed in these public forums are not endorsed by NCTM or The Math Forum.

Notice: We are no longer accepting new posts, but the forums will continue to be readable.

Topic: Hidden Markov Models with variable length sequences
Replies: 2   Last Post: Oct 29, 2013 12:45 AM

 Raul Posts: 1 Registered: 10/29/13
Re: Hidden Markov Models with variable length sequences
Posted: Oct 29, 2013 12:45 AM

Hi Sujai,

I am looking at how to use HMMs using Kevin Murphy's HMM toolbox and I came across the same issue, where in the example provided the length of the sequence (T) is fixed, and my data has variable length. You mentioned that you use cells, could you please upload an example of how to overcome variable length examples using this tool.

Thanks,
Raul

Sujai <NOsujaikumarSPAM@ameritech.net> wrote in message <bopj6h\$1hhptj\$1@ID-204957.news.uni-berlin.de>...
> > I can't remenber exactly the excellent Kevin Murphy's HMM toolbox but
> > I think you can use structure to take into account your problem. What
> > is exactly your problem ?. Is your number of state and measurement
> > symbol changing every time increment ?. In this case I thin you can
> > use the Kevin toolbox
> > with A and B two cell arrays of matrices. i.e.
> > A{1}(i,j) is equal to Pr(x_2=i|x_1=j), ...,
> > A{t}(i,j) is equal to Pr(x_{t+1}=i|x_{t}=j) etc..

>
> no, my problem was much simpler. i had many cases and each was of a diff
> length (i.e. the first data case had 5 observations, 2nd had 10 etc).
> since you mentioned cell arrays however, i looked those up and that
> seems to be exactly what i need. so thanks!
>
> - sujai
>
>