Jason Palmer

It seems that the sphering is being done on all the nodes, but the minimum eigenvalue is not being broadcast or something. The code may have been changed so that...

Hi Jazelle, Could you send the complete text output (on Matlab command line), which shows all the parameter settings? Thanks, Jason On Wed, Feb 13, 2019, 8:45 PM Jazelle759 Dear...

Hi Jazelle, For some reason Amica thinks there are 4 files of input data, instead of 1 as it should be. I’m not sure how this is happening, since the...

Hi Jazelle, The Files problem is gone, but now it seems there is problem with the data. The mean has NaNs. Could you check the data with any(isnan(EEG.data(:))). Also, the...

It seems to be using the .set file instead if the .fdt. Do you have the eeglab option set to save datasets in two files, .set and .fdt? Could you...

Ok, now it's running, but the data low rank is causing problems. Are you interpolating channels before ica? The best approach is to run ica on the data without interpolating...

Sorry you're having so many problems with amica. The covariance eigenvalues seem right though for avg ref. It should work if you use 'pcakeep', 123. The problem seems to be...

Could you try setting ‘max_threads’, 1 (instead of default 2) and see if it still sometimes dies? From: Jazelle759 [mailto:[email protected]] Sent: Wednesday, February 27, 2019 7:09 PM To: japalmer29/amica Cc:...

Maybe use fixed 'block_size' of 64, 128, or 256, with 'max_threads', 2. Maybe small enough block size won't crash and will be faster. On Wed, Feb 27, 2019, 9:36 PM...

Also use 'do_opt_block', 0, (or leave that keyword arg out). On Wed, Feb 27, 2019, 10:25 PM Jason Palmer wrote: > Maybe use fixed 'block_size' of 64, 128, or 256,...