Implementation of ATCnet

[cedricrommel]

No description provided.

[codecov]

Codecov Report

Merging #429 (6002542) into master (171e505) will increase coverage by 0.59%.
The diff coverage is 97.29%.

@@            Coverage Diff             @@
##           master     #429      +/-   ##
==========================================
+ Coverage   84.81%   85.40%   +0.59%     
==========================================
  Files          48       49       +1     
  Lines        3728     3913     +185     
==========================================
+ Hits         3162     3342     +180     
- Misses        566      571       +5     

[cedricrommel]

ping @tomMoral @agramfort :)

[agramfort]

LGTM although I would make all classes in ATCNet.py with a starting _ to be private except the ATCNet class that is public.

🙏 @cedricrommel

[cedricrommel]

LGTM although I would make all classes in ATCNet.py with a starting _ to be private except the ATCNet class that is public.

🙏 @cedricrommel

Thanks Alex! I didn't add the _ in some modules because I thought those were general enough to be useful in other future networks. I reckon I should have moved them into braindecode.models.modules :) I'll do it for those that still seem general enough and add the _ to the others.

[cedricrommel]

Feel free to merge if it looks good to you :)

[tomMoral]

LGTM.

Just one nitpick but this is on me, I don't like the warpping style when subclassing does the job. feel free to ignore.

[tomMoral]

Same can be done with the maxnorm layer.

[cedricrommel]

Same can be done with the maxnorm layer.

Good point! Did the same for MaxNormLiner