[Biojava-l] problems with the profilehmm demo

[Mark Schreiber]

Hi Daniel -

Your suggestion for the cause of the bug makes sense. I will take a
look at this tomorrow when I have access to the CVS.

A possible solution might be to make AlphabetIndexers listen to their
underlying Alphabets for changes to the number of Symbols.

Thanks for your help!  If you think of a solution before I wake up
tomorrow you can be biojava hero of the month : )

- Mark

On 8/29/07, daniel.rohrbach at web.de <daniel.rohrbach at web.de> wrote:
> Hi all, hi Mark
>
> thanks for the fast Answer.
>
> I looked at the code and perhaps solved the problem.
>
> During initialisation of the ProfileHMM ( bevore method connectModel() is called ) the transWeights are set with SimpleDistributions sd. These sd again are initialized with an empty Alphabet (no symbol). (*)
>
> These Distributions keeps an indexer Object which is used during the training process where the exception occurs.
>
> because of (*) the number of indexed symbols is 0 (indexer.symbols.length == 0) and the (FiniteAlphabet)indexer.alphaRef.get() has length 0 too.
> (Indexer is of type LinearAlphabetIndex)
>
> When connectModel() is called the Alphabets ( which are the same as indexer.alphaRef ) are filled with some symbols (EmissionState, DotState ...) (**) BUT the indexer.symbols.length will not be changed (***).
>
> This causes the following exception:
> In SimpleDistribution.Trainer.trainImpl is a method called getWeights(). This will return a double list of length indexer.getAlphabet() which is the same of indexer.alphaRef. Because of (**) this will have a length n>0. So that the for loop in this piece of code runs from 0 to n. Here the exception occurs. indexer.symbolForIndex(i) (0<=i<n) cause the error because of (***) indexer.symbols are empty.
>
> I think it must be changed the symbols of the indexer Object in the connectModel() method too. But I'm not really sure because I didn't worked with biojava so long.
>
> I'm sorry for my bad English but I hope that I could help a little.
> Many greetings
> Daniel
>
> > -----Ursprüngliche Nachricht-----
> > Von: "Mark Schreiber" <markjschreiber at gmail.com>
> > Gesendet: 28.08.07 16:28:23
> > An: "Daniel Rohrbach" <daniel.rohrbach at web.de>
> > CC: biojava-l at lists.open-bio.org
> > Betreff: Re: [Biojava-l] problems with the profilehmm demo
>
>
> >
> > Hi -
> >
> > This is a bug that has been submitted to bugzilla. Strangely it was
> > not an issue in the pre-releases of biojava.  I was kind of hoping
> > someone else would fix it but I guess I will have to find time to take
> > a look.
> >
> > - Mark
> >
> > On 8/28/07, Daniel Rohrbach <daniel.rohrbach at web.de> wrote:
> > > Hi all,
> > >
> > > i'm trying to run the  "How do I make a ProfileHMM" example but it doesn't want to work.  I copied the code to my program and when I start it I get the following exceptions:
> > >
> > > Exception in thread "main" java.lang.IndexOutOfBoundsException: Can't find symbol for index 0
> > >         at org.biojava.bio.symbol.LinearAlphabetIndex.symbolForIndex(LinearAlphabetIndex.java:106)
> > >         at org.biojava.bio.dist.SimpleDistribution$Trainer.trainImpl(SimpleDistribution.java:332)
> > >         at org.biojava.bio.dist.SimpleDistribution$Trainer.train(SimpleDistribution.java:306)
> > >         at org.biojava.bio.dist.SimpleDistributionTrainerContext.train(SimpleDistributionTrainerContext.java:126)
> > >         at biojavatests.HMMTest.main(HMMTest.java:71)
> > >
> > > The error occurs at mt.train(). It seems that there was no, wrong or empty alphabet set for the trainer that shall be used.
> > >
> > > here the link to the examplepage: http://biojava.org/wiki/BioJava:CookBook:DP:HMM
> > > any help on this topic?
> > >
> > > Thanks in advance
> > > __________________________________________________________________________
> > > Erweitern Sie FreeMail zu einem noch leistungsstärkeren E-Mail-Postfach!
> > > Mehr Infos unter http://produkte.web.de/club/?mc=021131
> > >
> > >
> > > _______________________________________________
> > > Biojava-l mailing list  -  Biojava-l at lists.open-bio.org
> > > http://lists.open-bio.org/mailman/listinfo/biojava-l
> > >
> >
>
>
> _______________________________________________________________________
> Jetzt neu! Schützen Sie Ihren PC mit McAfee und WEB.DE. 3 Monate
> kostenlos testen. http://www.pc-sicherheit.web.de/startseite/?mc=022220
>
>
> _______________________________________________
> Biojava-l mailing list  -  Biojava-l at lists.open-bio.org
> http://lists.open-bio.org/mailman/listinfo/biojava-l
>