[Eeglablist] ICA calculation on core i7 vs. i9

[Makoto Miyakoshi]

Dear Arno,

> I would be careful about spending up ICA by simply reducing the number of
samples. The number of sample is important in reaching a stable solution
(the more independent observations, the better).

If simply the number of datapoint were to matter, it would be possible to
record 1-second 128ch EEG data at 1MHz sampling and run ICA to obtain
satisfactory results. But apparently it would not work.
When I discussed this issue with Luca, he suggested that we should rather
consider the length of data length in second rather than the number of
datapoints. I found his argument convincing. Hence I always recommend that
our convention of 20-30*ch^2 applies at 250 HZ.

> This would deserve numerical comparisons.

I totally agree with you.This kind of validational study is necessary and
would make a good low-hanging-fruit publication for grad students.

Makoto

On Sun, Jul 19, 2020 at 4:49 PM Delorme, Arnaud <adelorme at ucsd.edu> wrote:

> Dear Hadar and Makoto,
>
> I would be careful about spending up ICA by simply reducing the number of
> samples. The number of sample is important in reaching a stable solution
> (the more independent observations, the better). This would deserve
> numerical comparisons.
>
> Also, there is an implantation of Infomax on GPU (CudaICA) at the bottom
> of this page as well as a Matlab interface for it.
>
> https://sccn.ucsd.edu/wiki/Binica
>
> Arno
>
> > On Jul 18, 2020, at 1:09 PM, Makoto Miyakoshi <mmiyakoshi at ucsd.edu>
> wrote:
> >
> > Dear Hadar,
> >
> > If you are talking about runica(), single core clock speed matters. I
> don't
> > think it has parallel computing capabilities.
> >
> >> P.S. any other (mainly hardware) recommendations to speed up eeglab's
> > calculations would be welcome!
> >
> > I added a new section for you in the following Wiki page.
> >
> https://sccn.ucsd.edu/wiki/Makoto's_useful_EEGLAB_code#How_to_speed_up_ICA_.2807.2F18.2F2020_added.29
> >
> >
> > Makoto
> >
> >
> > On Tue, Jul 14, 2020 at 2:41 AM Hadar Levi Aharoni <
> > hadar.levi at mail.huji.ac.il> wrote:
> >
> >> Hi EEGLAB list
> >>
> >> We are currently considering whether to buy a pc with a core i7 or i9,
> but
> >> we're not sure whether it's worth the cost in terms of speed up.
> >> I am going to do a lot of ICA calculations, so I was wondering if anyone
> >> here has any information regarding eeglab's ICA performance on i9 vs.
> i7?
> >> Is there a big difference in the calculation time?
> >>
> >> P.S. any other (mainly hardware) recommendations to speed up eeglab's
> >> calculations would be welcome!
> >>
> >> Many thanks,
> >>
> >> Hadar
> >> _______________________________________________
> >> 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
> >>
> > _______________________________________________
> > 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
>
> _______________________________________________
> 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
>