[Pkg-exppsy-maintainers] Dataset label question

[Yaroslav Halchenko]

> I've a need to have more complex labels, or at least supplementary
> labels, for a dataset.  I tried passing in an array with two columns
> for the labels, but this does not work because the unique values
> fails. 
I am in need myself I think for some 'multidimensional labels'... I
haven't thought about it yet though...

What could help you out in such situation is the fact that labels as
well as chunks are added into dataset class via... let me just cite:

# Following attributes adherent to the basic dataset
Dataset._registerAttribute("samples", "_data", hasunique=False)
Dataset._registerAttribute("labels",  "_data", hasunique=True)
Dataset._registerAttribute("chunks",  "_data", hasunique=True)

so... you could extend default Dataset or better subclass it and define
some Dataset to which you add new attribute without unique value
computation assigned (if that is needed).

Would that help?

> What I need is that when the dataset is split for cross
> validation based on the chunks, I need to be able to perform some
> additional selectSamples in the train_dat, but the labels, themselves,
> are not enough information for me because I need to remove some
> samples, but only if they are not of a specific type (that is not
> defined by the labels.)
I think that above-mentioned could resolve it for you, right?


> Any ideas?  Is there anything already in place?

> Thanks,
> Per

> _______________________________________________
> Pkg-exppsy-maintainers mailing list
> Pkg-exppsy-maintainers at lists.alioth.debian.org
> http://lists.alioth.debian.org/mailman/listinfo/pkg-exppsy-maintainers


-- 
Yaroslav Halchenko
Research Assistant, Psychology Department, Rutgers-Newark
Student  Ph.D. @ CS Dept. NJIT
Office: (973) 353-5440x263 | FWD: 82823 | Fax: (973) 353-1171
        101 Warren Str, Smith Hall, Rm 4-105, Newark NJ 07102
WWW:     http://www.linkedin.com/in/yarik