[Eeglablist] extremely slow ICA

[Makoto Miyakoshi]

Dear Brittany,

The trimOutlier() plugin can let you 'trim' the outliers i.e., cut out the
window for rejection. Some people say this is simple and good way to go to
'clean' your data. However, if you have so many single-channel artifacts,
because this method will reject a chunk across all the channels, you may
end up with too much data rejection. If this is an issue, use
clean_rawdata() plugin. This plugin as artifact subspace rejection (ASR)
algorithm. For more info, see the link below.
https://sccn.ucsd.edu/wiki/Makoto's_preprocessing_pipeline#Alternatively.2C_cleaning_continuous_data_using_ASR_.2803.2F16.2F2018_updated.29

I like ASR because it is less wasteful. For trying it out, start with SD ==
20 for ASR parameter ('Repair burst' part).

Makoto

On Fri, Jul 6, 2018 at 12:05 PM Brittany Alperin <balperin07 at gmail.com>
wrote:

> Thanks Makato
>
> If I do find outliers, is there anything I can do about it? Any idea what
> types of things would cause this in the data?
>
> Thanks,
> Brittany
>
>
> On Fri, Jul 6, 2018 at 11:57 AM, Makoto Miyakoshi <mmiyakoshi at ucsd.edu>
> wrote:
>
>> Dear Brittany,
>>
>> The speed of Infomax implemented in runica() does get affected by data
>> quality.
>> I recommend you check data data set using trimOutlier() plugin.
>> https://sccn.ucsd.edu/wiki/TrimOutlier
>> Go back to the continuous data and run trimOutlier. It will show you
>> envelope of all channels across time. Well-tempered data should show
>> 'stationary' (i.e., the envelope shapes more or less rectangular with
>> constant width) shape. If you have super huge outlier or continuous zeros,
>> you will easily find it. Such outliers can make the infomax really slow.
>>
>> Makoto
>>
>> On Fri, Jul 6, 2018 at 10:31 AM Brittany Alperin <balperin07 at gmail.com>
>> wrote:
>>
>>> Hello
>>>
>>> I'm currently working with a fairly large dataset and the ICA takes
>>> several hours to run, which I consider normal. I have one file, which is
>>> seemingly no different from the others, but the ICA has been running for
>>> ~40 hours and it's not even half way done.
>>>
>>> I've had this happen with different datasets before, but it's fairly
>>> rare so I never thought much of it until now.
>>>
>>> Has anyone had this happen? Any thoughts? I'm using EEGLAB 12.0.2.6b,
>>> Matlab 2012b, and a Windows 7 Enterprise OS.
>>>
>>> Thanks,
>>> Brittany
>>>
>>> *Brittany Alperin*
>>>
>>> PhD Candidate
>>>
>>> ATTEND LAB
>>>
>>> Department of Behavioral Neuroscience
>>>
>>> Oregon Health & Science University
>>>
>>> Portland, OR
>>>
>>> 503-418-8402
>>>
>>> _______________________________________________
>>> 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
>>
>>
>>
>> --
>> Makoto Miyakoshi
>> Swartz Center for Computational Neuroscience
>> Institute for Neural Computation, University of California San Diego
>>
>
>

-- 
Makoto Miyakoshi
Swartz Center for Computational Neuroscience
Institute for Neural Computation, University of California San Diego
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20180706/6f60eecb/attachment.html>