Re: Murphy HMM problem


John D'Errico wrote:
In article <1146736031.071137.324230@xxxxxxxxxxxxxxxxxxxxxxxxxxxx>, "PaulN" <paulnicholl@xxxxxxxxxxx> wrote:

I'm using Kevin Murphy's HMM toolbox to model faces. However, when I
run the Expectation Maximisation function - mhmm_em - I'm getting the
following error:

??? Error using ==> assert
assertion violated:

Error in ==> fwdback at 150
assert(approxeq(sum(alpha(:,t)),1));

Error in ==> mhmm_em>ess_mhmm at 145
[alpha, beta, gamma, current_loglik, xi, gamma2] = ...

Error in ==> mhmm_em at 69
[loglik, exp_num_trans, exp_num_visits1, postmix, m, ip, op] = ...

Error in ==> trainSingleHMM at 20
[LL, prior1, transmat1, mu1, Sigma1, mixmat1] =...
...

Has anyone encountered this error before? I'm a bit of a novice when
it comes to HMMs, so I'm not really sure what's going on. My HMMs have
5 states, and each Gaussian mix has 5 mixtures.

Thanks,
Paul.

Only a guess, but try this in matlab

which assert -all

Do you have multiple functions named assert.m on
your search path? Have you downloaded the function
assert.m from the file exchange?

HTH,
John D'Errico


--
The best material model of a cat is another, or preferably the same, cat.
A. Rosenblueth, Philosophy of Science, 1945

Those who can't laugh at themselves leave the job to others.
Anonymous

Thanks John.

The function assert just checks (in this case) that the current column
in the alpha matrix equals 1, and produces an error if it doesn't.
Unfortunately, I am getting columns in my alpha matrix (calculating the
forward variables) whose sum is zero. Maybe somebody who has used
Murphy's toolbox might know why this could come about.

.

Relevant Pages

  • Re: Compile Time assertions
    ... You asserted something at compile time, ... >> the assumption and you got a compiler error. ... > Yeah John... ... > But am i not supposed to assert when the pointer is null. ...
    (comp.lang.cpp)
  • Re: religion a function of race
    ... John of Aix wrote: ... It's been heading that way for a long time, and no, I don't assert ... I saw evidence ...
    (uk.politics.misc)
  • Re: Fastcode GCD
    ... is it pointless using assert (most programmers ... > don't/won't recompile the RTL). ... Assert is out of the question and we vote about Exception ... John and Aleksandr and Hroje against. ...
    (borland.public.delphi.language.basm)
  • Re: Anyone use ELSE minor-mode in Emacs?
    ... [John J Lee] ... :-) I tried Pythonwin a bit, ... it would be presumptuous for me to assert anything about it. ...
    (comp.lang.python)
  • Re: Incompleteness vs. Mechanical Reasoning
    ... the decision (in which case at least X can correctly assert the truth ... And this is exactly what NAFL rejects. ... An analogiy with quantum physics is in the Schrodinger cat example. ... Until that time the cat remains in a superposed state. ...
    (sci.logic)
Loading