Dimensions problems with using searchlight

✉️ Mailing List Archive (read-only)
1

Hi MNE,

I am trying to apply the example of using CSP to decode for motor responses
and in time frequency, but I keep getting an error from my scores =
cross_val_multiscore(time_decod, X, y, cv=5, n_jobs=1), saying "X must have
at least 3 dimensions." I know that the X that i put in is the 3-D epochs
data, but this error comes from a command run in the CSP function with
"def _check_Xy(self, X, y=None):" I was wondering what I could do to solve
this issue.

Aditya Singh
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20171024/423c2bb4/attachment.html

2

Your script suggests that your estimator (time_decod) is a
SlidingEstimator. Since you mention using CSP, I guess you did
time_decod = SlidingEstimator(CSP())

If so, SlidingEstimator will loop over the last dimension of X. If your X
is of shape (n_epochs, n_channels, n_times), then it will loop over time.

This means that CSP only has access to a single time sample, hence the
complain.

Hope that helps,

JR