[pymvpa] Invariant features and hyperalignment

Beau R. Sievers Beau.R.Sievers.GR at dartmouth.edu
Tue Jul 17 16:38:07 BST 2018

Hi PyMVPA world,

I am attempting searchlight hyperalignment, and I’m running into a problem with invariant features.

If I do _not_ remove invariant features from my datasets, I get an error when attempting to use an instance of SearchlightHyperalignment to calculate the maps: "ValueError: For now do not handle invariant in time datasets”

If I _do_ remove invariant features, I am able to calculate the maps, but when I call the .forward() method to map a dataset into the common space, I get a different error: "ValueError: dimension mismatch.” I presume (but cannot be sure) that this is related to the datasets now having different numbers of features.

Does anybody have experience with this issue or know what I might be doing wrong? This is using PyMVPA 2.6.3.dev1 (as installed on the “social cluster” at Dartmouth).

Thanks for your help,

More information about the Pkg-ExpPsy-PyMVPA mailing list