[pymvpa] Hyperalignment: SVD did not converge

Yaroslav Halchenko debian at onerussian.com
Tue May 1 01:17:44 UTC 2012

Hi Kiefer,

thanks for the investigation.  First let's clarify some 'terminology' ;)

> From your previous response in a different thread you mentioned that
> SVD does not converge when it is unable to find an appropriate dataset
> to kick off hyperalignment. This doesn't quite make sense to me ...

to me neither ;)  it is somewhat the opposite -- Hyperalignment uses
Procrustean transformation which relies on SVD.  If SVD doesn't converge
(having some subject as a tentative target space) then Procrustean fails
and then Hyperalignment tries another subject as the target.

> Could the SVD non-convergence be a classification error on my part?

"Classification error"... ?  we are not talking about classification at
this stage yet

> Out of the 21 subjects, only a group of 6 subjects managed to
> successfully hyperalign to each other by the end.

hm... so let's talk now about details -- how many time points do you
use? and how many voxels do you hyperalign? how do you select those

my wild guess is that you have way too many voxels for the # of time
points you have... 

Keep in touch                                     www.onerussian.com
Yaroslav Halchenko                 www.ohloh.net/accounts/yarikoptic

More information about the Pkg-ExpPsy-PyMVPA mailing list