[Biojava-l] viterbi training in biojava

[wendy wong]

thanks! i only need one head so BaumWelchSampler works fine with me.
The default SCORETYPE is probability and when I tried it the score
goes back and forth, like + for one time and - for the next time. I
then changed it to LOGODDS  and recompiled biojava and now that the
score is steadily increasing. I was wondering if the SCORETYPE could
be passed in as an argument in the next version of biojava?

thanks,
wendy


On 30 May 2006 12:19:15 +0100, David Huen <smh1008 at cam.ac.uk> wrote:
> On May 30 2006, wendy wong wrote:
>
> >Hi,
> >
> >I was wondering if viterbi training is implemented in biojava, or if
> >there's any open source version implemented using biojava?
> >
> There is one-head viterbi training already I think. The training framework
> doesn't work for two-head - I wrote a viterbi training API that works for
> two head but it is not fully compatible with the existing API so I never
> put it into CVS, plus it doesn't have Baum-Welch implemented either.
>
> If it is any use to you you can have it.
>
> Regards,
> David
>