AH Ah ah..... THANKS -- I should have thought to grep for it Swaroop -- could you would you in the plane write a little essay on why would we want to use dgesvd as the default for Hyperalignment?
We might like then add 'auto' setting for svd which would choose dgesvd if available and regular numpy's svd if not... Ideally imho I would prefer to have numpy's implementation as the default if in 99% of the cases they provide comparable results at comparable speeds and dgesvd as the workaround for some tricky cases. ok, Kiefer -- quick workaround would be to initialize your Hyperalignment with alignment=Procrustean() without custom svd settings. On Fri, 27 Apr 2012, Kiefer Katovich wrote: > I did not at any point in my script manually set svd='dgesvd' > However, in hyperalignment.py at line 47 this is automatically set: > # Lets use built-in facilities to specify parameters which > # constructor should accept > alignment = Parameter(ProcrusteanMapper(svd='dgesvd'), # might > provide allowedtype > doc="""The multidimensional transformation mapper. If > `None` (default) an instance of > :class:`~mvpa2.mappers.procrustean.ProcrusteanMapper` is > used.""") > _______________________________________________ > Pkg-ExpPsy-PyMVPA mailing list > [email protected] > http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/pkg-exppsy-pymvpa -- =------------------------------------------------------------------= Keep in touch www.onerussian.com Yaroslav Halchenko www.ohloh.net/accounts/yarikoptic _______________________________________________ Pkg-ExpPsy-PyMVPA mailing list [email protected] http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/pkg-exppsy-pymvpa
