home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!munnari.oz.au!manuel.anu.edu.au!andosl!tridge
- From: tridge@andosl.anu.edu.au (Andrew Tridgell)
- Newsgroups: comp.speech
- Subject: Re: Null Transitions in HMMs
- Date: 21 Jan 1993 22:14:45 GMT
- Organization: CSLab, Autralian National Uni.
- Lines: 51
- Distribution: world
- Message-ID: <1jn78mINNbhm@manuel.anu.edu.au>
- References: <1993Jan20.031204.1652@seas.gwu.edu>
- NNTP-Posting-Host: 150.203.15.21
-
- In article <1993Jan20.031204.1652@seas.gwu.edu>, marshall@seas.gwu.edu (Christopher Marshall) writes:
- > Back in October I posted a plea for help in understanding null transitions in
- > HMMs. My thanks to Imoto Takashi, Les Niles, and Wieland Eckert who spent a
- > lot of time corresponding with me through email to try to explain them to me.
- >
- > Unfortunately, although I learned some things, I never learned how
- > to implement null transitions in the evaluation problem of HMMs. I have
- > since left and come back to the problem.
- > etc....
-
-
- > Now, I can not for the life of me figure out how to modify this procedure
- > to incorporate null transitions.
- >
- > Does anyone know how to do this?
-
- The solution to this is to NOT change the equations at all - but to
- change A(i,j) and B(i,j,k).
-
- A(i,j) is defined by you as the probability of moving from state i to
- state j in 1 time step. If we instead define it as the probability of
- moving from state i to state j while producing exactly one output then
- the problem is solved. If there are no null transitions then these
- definitions are equivalent.
-
- B(i,j,k) is similarly defined normally as the probability of producing
- symbol k given that a transition is made in one time step from state i
- to state j. The new definition will be the probability of producing
- symbol k given that exactly one output is produced in a move from
- state i to state j.
-
- An example will make this clearer (I hope). Say you already have the
- A(i,j) and B(i,j,k) for no null transitions. Then say we add a null
- transition from state s1 to state s2 and wish to find A'(i,j) and
- B'(i,j,k) including this transition. We can write:
-
- A'(i,s2) = A(i,s2) + A(i,s1)
- B'(i,s2,k) = B(i,s2,k) + B(i,s1,k)
-
-
- I will be interested to hear of any other suggestions you may receive.
-
-
- Andrew
-
-
- --
- =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
- Andrew Tridgell CSLab, Research School of Physical Sciences
- Andrew.Tridgell@anu.edu.au Australian National University (x3064)
- =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
-