Open F-said opened 3 years ago
So it appears that autoreject splits the data into K folds (defaulted to 10) and within each fold you have a train set and a test set.
WITHIN THE TRAIN SET OF EACH FOLD: """ apply threshold τi to reject trials in the train set calculate the mean of the signal (for each sensor and timepoint) over the GOOD (=not rejected) trials in the train set """ WITHIN THE TEST SET OF EACH FOLD: "calculate the median of the signal (for each sensor and timepoint) over ALL trials in the test set"
WITHIN EACH FOLD: "compare both of these signals and calculate the error ek (i.e., take the Frobenius norm of their difference)"
What I THINK is happening:
If when we apply threshold τi to the train set and we end up rejecting all channels, we aren't able to compute the mean of non-rejected channels and therefore can't find an error for that fold, and I assume how the code is handling this on the autoreject side is to have rejected vs non-rejected channels be marked by the label variable y, a similar error should occur if all channels are accepted if this is correct.
I'll keep looking into this.
Describe the bug For the
sub-NDARAB793GL3_ses-01_task-Video1_run-01_eeg
file, the following error occurs duringfinal_reject
Expected behavior No crashes should be expected during running of pipeline
Screenshots Entire error message is as follows: