[Eeglablist] EEGLAB ASR Question

[Makoto Miyakoshi]

Dear John and Bennett,

I updated clean_rawdata() to ver. 1.00 in which using 'off' in GUI is now
officially supported. Of course, in clean_rawdata() 'off' is supported. If
you find a problem, please let me know.

I also separated some descriptions about ASR/clean_rawdata() to a separate
Wiki page in which I put a summary report of the output data length
fluctuation issue and the solution. In short, always use 'availableRAM_GB'
option, otherwise clean_rawdata() does not guarantee complete
replicatability.
https://sccn.ucsd.edu/wiki/Artifact_Subspace_Reconstruction_(ASR)

Makoto

On Mon, Apr 1, 2019 at 5:08 PM Johnson, John T. <john.johnson at gatech.edu>
wrote:

> Using -1 is the correct way to disable options.
> Using the text off may cause undefined behavior. With the conversion from
> string to number in processing the GUI controls, using off for a
> parameter would result in that parameter being set to zero. This may be the
> source of the odd behavior Bennett was seeing.
>
> Thanks for your reply,
> John
>
> On 18 Mar 2019, at 22:13, Makoto Miyakoshi wrote:
>
> Dear Bennett and John,
>
> Sorry to hear the trouble. I'm interested in replicating the problem.
> Could you please tell me more details to replicate the issue? Let me see if
> I can fix it. Please make sure you are using EEGLAB 14. Up to ver.13 there
> was a quite serious bug that is related to handling events.
>
> However, while manually running ASR in the GUI does not result in the
>
> loss of events, running ASR from code does (and -1 vs. 'off' yield
> different numbers of events lost). Is there any way to resolve this issue?
>
> I'm not sure about 'event lost' unfortunately, but 'different number of
> events lost' may be related to this: Every time ASR is performed, it first
> obtains the available RAM. Try the following code on MATLAB.
>
> freeRamInMB
> =
> java.lang.management.ManagementFactory.getOperatingSystemMXBean().getFreePhysicalMemorySize()/(2^20);
>
> This tells you how much RAM is available, seen from this command. ASR
> determines how it divide the continuous data into chunks, which can
> actually change the final data size (my colleague told me up to 0.3% of
> jitter was confirmed.) If you are curious, change the code in asr_process
> line 105 maxmem = XXX (this should be MB*2). This should fix the number of
> 'rejected events.'
>
> Now, general discussion... If we want to obtain the complete
> reproducibility, then we have to hard-code this value. ASR was originally
> written for real-time BCI application, so this kind of solution was
> necessary to prevent memory overflow. However, for the offline analysis we
> may want to use a fixed value. However, the current plugin does not have a
> structure to pass the value down to this function. I will modify the code
> with my colleagues' help (hopefully including Christian himself) to address
> the issue. It may take some time (I learned this thing in last November!),
> but I want to address it by taking advantage of this momentum!
>
> Makoto
>
>
>
> On Thu, Mar 14, 2019 at 9:18 AM Alterman, Bennett L <balterman3 at gatech.edu
> >
> wrote:
>
> If using the GUI for ASR, using -1 to turn off an input field yields the
> loss of event markers (and eventually epochs), but writing off in the
> fields does not result in this loss. When using eegh to get the code, it
> returns a line that uses off, but the code does not run unless 'off' is
> used. However, while manually running ASR in the GUI does not result in the
> loss of events, running ASR from code does (and -1 vs. 'off' yield
> different numbers of events lost). Is there any way to resolve this issue?
>
>
> Sincerely,
> Bennett Alterman, M.S.
>
> --
> Ph.D. Student, Cognitive Motor Control Laboratory
> School of Biological Sciences
> Georgia Institute of Technology
> bennett.alterman at gatech.edu
> http://bennettalterman.wix.com/neuroimaging
> <http://bennettalterman.wixsite.com/neuroimaging>
> https://www.linkedin.com/in/bennettalterman/
> 678.485.6318
>
> _______________________________________________
> 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
> Assistant Project Scientist, Swartz Center for Computational Neuroscience
> Institute for Neural Computation, University of California San Diego
>
> John T. Johnson
> PhD Candidate - Cognitive Motor Control Laboratory
> Lab TA NEURO 2001 Principles
> School of Biological Sciences
> Georgia Institute of Technology
>
> 678-575-2093
> john.johnson at gatech.edu
>


-- 
Makoto Miyakoshi
Assistant Project Scientist, 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/20190401/b4c4bef7/attachment.html>