[Eeglablist] AMICA error on 256 channels EEG

[liao ke]

Hi Jason,

Thank you for your reply!

I tried to use fixed blocksize of 64 and 32 and the error stays the same. I
also tested AMICA  can run on Linux on my data followed
Hyeonseok's solution. Thanks for your help!

Best,
Ke

On Sat, Mar 19, 2022 at 2:49 PM Jason Palmer <japalmer29 at gmail.com> wrote:

> Hi Ke,
>
> You might try reducing the blocksize. Use fixed blocksize of 64 or 32.
> Default is either to optimize the blocksize or use fixed size of 256 I
> think. If there is a stack overflow the program can do a segmentation fault
> and die.
>
> Best,
> Jaskn
>
> On Thu, Mar 17, 2022, 4:19 PM liao ke <kekeliao at gmail.com> wrote:
>
>> Hi Jason,
>>
>> Thank you for your reply!
>>
>> Here are the minimum eigenvalues in the out.txt file of a 256-channel EEG
>> dataset with AMICA running error:
>>
>>  minimum eigenvalues =   0.516290976112914       0.635721497064936
>>   0.721411922551856
>>  maximum eigenvalues =    8727729.20865751        1896971.09878320
>>    425608.476260537
>>  num eigs kept =          256
>>
>> I am not sure if the condition number is high. Or how to set mineig
>> keyword. When set PCA as 180 from GUI, AMICA can run on this dataset. I
>> also used the code from Makoto EEGlab webpage to test the data rank :
>> dataRank = sum(eig(cov(double(EEG.data'))) > 1E-7)
>> dataRank = rank(double(EEG.data'))
>> From both code the data rank are 256.
>>
>> Best,
>> Ke
>>
>> On Fri, Mar 11, 2022 at 12:31 PM Jason Palmer <japalmer29 at gmail.com>
>> wrote:
>>
>>> Hi Ke,
>>>
>>> Could you check the output directory for and out.txt file for the 256
>>> channel runs?
>>>
>>> It may be that the data rank is much lower than the number of channels.
>>> You can check the lowest eigenvalues of the data with
>>> eig(cov(EEG.data(:,:)')). The mineig keyword option can be used to set the
>>> level of the minimum eigenvalue size to keep dimensions in PCA. Or you can
>>> set the number of dimensions to keep with the pcakeep keyword option.
>>>
>>> Best,
>>> Jason
>>>
>>> On Fri, Mar 11, 2022, 9:01 AM liao ke via eeglablist <
>>> eeglablist at sccn.ucsd.edu> wrote:
>>>
>>>> Dear all,
>>>>
>>>> I have a problem in running the AMICA and need your help.
>>>>
>>>> I have used AMICA to process some 128 channel EEG datasets recorded from
>>>> EGI system (reduced to 64 channels for calculation)  without any
>>>> problem.
>>>> When I run AMICA on some 256 channel EEG data from the same system, it
>>>> always gives me the error message like:
>>>> 'No gm present, setting num_models to 1
>>>> No W present, exiting
>>>> Reference to non-existent field 'W'.
>>>>
>>>> Error in runamica15 (line 895)
>>>>     weights = mods.W(:,:,1);
>>>>
>>>> Error in pop_runamica (line 338)
>>>>     [W,S,mods] = runamica15(g.datfile,arglist{:});
>>>>
>>>> Error while evaluating Menu Callback.'
>>>>
>>>> This happens even I reduce the data into 183 channels. But when I
>>>> further
>>>> reduce the data into 90 channels, the AMICA can run well. I am using
>>>> AMICA
>>>> v1.6.1 with all default parameters from GUI.
>>>>
>>>> Is there any setting wrong with  my AMICA? Is there a channel limitation
>>>> with AMICA?
>>>>
>>>> Thanks very much for your help!
>>>> Ke
>>>> _______________________________________________
>>>> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
>>>> To unsubscribe, send an empty email to
>>>> eeglablist-unsubscribe at sccn.ucsd.edu
>>>> For digest mode, send an email with the subject "set digest mime" to
>>>> eeglablist-request at sccn.ucsd.edu
>>>>
>>>