Clemens Brunner comments

Results 317 comments of


Clemens Brunner

Use np.linalg.multi_dot instead of multiple np.dot routines

If we're already benchmarking, if it turns out that `@` is not significantly slower, I'd vote for that option because it's the best in terms of readability.

Use np.linalg.multi_dot instead of multiple np.dot routines

Likely, but it would be better to check.

Use np.linalg.multi_dot instead of multiple np.dot routines

On my computer, `multi_dot` always performs worst, and `@` best. In addition, `@` is also best in readability, so my vote goes to replacing everything with `@`. ``` # multi_dot,...

Use np.linalg.multi_dot instead of multiple np.dot routines

OT: I'm a bit disappointed. Your M1 is only about 2x faster than my 8 year old MacBook Pro: ``` multi_dot dot @ shapes [(4, 4), (4, 1), (1, 4)]...

Use np.linalg.multi_dot instead of multiple np.dot routines

With everything I've read about the M1, it should not throttle at all while on battery, and even the slowest one should be much faster than an 8 year old...

Use np.linalg.multi_dot instead of multiple np.dot routines

Maybe something to report upstream? I thought the only reason why `multi_dot` exists is because it is faster...

Use np.linalg.multi_dot instead of multiple np.dot routines

Just a quick final question: what is the optimal order? Why do large matrices have to come last and what makes this faster than any other order?

EEGLAB file export data type issue

It could be argued that this is an issue with MATLAB. The number of trials is an integer number. Do integers need to be converted to floats when performing division...

Addition of a non-projection-based method of removing EOG and ECG artifacts.

I'd add the "standard" regression approach first, because that's widely used (and often requested). I have implemented one of the first regression-based approached described in Hillyard & Galambos (1970) [here](https://cbrnr.github.io/posts/removing-eog-regression/)....

Addition of a non-projection-based method of removing EOG and ECG artifacts.

I think so too! It would be nice if that function had a parameter to choose from Croft & Barry and Hillyard & Galambos (since the latter is very often...