[Eeglablist] Time-frequency transform became 150-450% faster (without using parallel computing toolbox)

[Delorme, Arnaud]

Hi Makoto,

Did you check the code of the timefreq function which is being used by the newtimef function of EEGLAB to compute time-frequency decompositions?
There are many different variants (some commented). Also, when using more than one channel, the time-frequency decomposition of all channels is computed together using a single matrix multiplication.  If you think your solution might still be useful, it would be great to add it to the timefreq function.

Cheers,

Arno

> On Nov 29, 2020, at 2:25 PM, Makoto Miyakoshi <mmiyakoshi at ucsd.edu> wrote:
> 
> Dear EEGLAB users,
> 
> I ran an experiment to test if removing for-loops can speed up
> time-frequency transformation. For epoched data, I confirmed 150%-200% of
> processing speed. The more time/frequency point there are, the more
> efficient it became. The most noticeable effect was found when I processed
> a continuous data with it. I confirmed speed increase to 450%. If you are
> interested in trying it out, please download the modified timefreq.m from
> the following link.
> 
> https://sccn.ucsd.edu/wiki/Makoto's_useful_EEGLAB_code#How_to_perform_time-frequency_transformation_faster_.2811.2F29.2F2020_added.29
> 
> This is a beta test. When you encounter a bug or problems, please let me
> know.
> 
> Makoto
> _______________________________________________
> 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