[pymvpa] libsvm dense arrays
Scott Gorlin
gorlins at MIT.EDU
Tue Sep 30 23:25:06 UTC 2008
> if you are using released version of pymvpa, then you can do dirty hack.
>
> prior to any mvpa imports
>
> import mvpa.base.externals
> mvpa.base.externals._VERIFIED['libsvm'] = False
>
> so it would trigger pymvpa to say that libsvm is not available.
>
> Please let us know how it works for you
>
>
This works quite well, everything is switched to Shogun and runs quite
nicely.
But it doesn't seem to matter - at least with some preliminary tests
(looking at model selection, so the code is going back and forth between
python<->libsvm/shogun quite a bit) using Shogun nets exactly the same
run times. More worrisome is that I get worse cross-validation
accuracies using Shogun for gamma < 0.5 in an Rbf classifier, though the
results seem comparable otherwise.
I just switched from the ubuntu Shogun package (0.4) to the current one
(0.6), so now I'm more confused than ever :)
If time performance is the same, does that indicate Shogun does *not*
use a dense data representation?
If model selection yields different optimal choices due to (sometimes)
different cross-validation rates, doesn't that suggest that libsvm and
shogun are using fairly different code bases?
by no means have I tested this on a robust dataset, but this simply
doesn't feel right :)
More information about the Pkg-ExpPsy-PyMVPA
mailing list