GSoC Idea: Extending MNE-python decoding module

đź’¬ Support & Discussions
1

Hi All,

I am computer science, joint Ph.D. student at University of Groningen, Netherlands and Indian Institute of Techology, Roorkee, India and would like to apply for GSoC this time. Till now, I have worked on broad fields of AI ranging from symbolic (cognitive modelling) to connectionist (deep learning). You can find more about me here .
Unfortunately I am sort of a newbie in open source contributions and would really like to contribute to MNE-python. My idea would be to extend MNE’s decoding module to include suitable deep learning algorithms. I know this is sort of vague right now, but do let me know if this idea piques your interest and we can talk more about the specifics.

Thank you.

2

Thanks for your interest! If you’re “a newbie in open source contributions” you could start by reading our (fairly extensive) contributor’s guide. It walks through how to set up a local environment capable of building our docs and running our tests.

Given that your research involves EEG I’m assuming you’ve already used MNE-Python at least a little bit (?) but there’s probably something in our docs you’ve never done, so studying our tutorials and examples might help you understand better the full scope of our software, and might help you clarify what is indeed a somewhat vague proposal so far.

Finally, I encourage you to take a look at our open issues (especially the 1.0 release milestone) and see if there are any issues you feel comfortable tackling.

2 Likes
3

Hi Anmol,
thanks for proposing a GSoC idea. When it comes to deep learning in EEG research there is already a mne-associated package for that, braindecode, so if you are interested in this area it might be worth considering focusing your efforts there. @agramfort might be able to tell you more. Another nice contribution could be to translate the ICLabel network from matlab to python (see this post for links to the repo and paper and more context).
But, just as @drammock notes, it is always a good idea to start getting familiar with open source contributions by fixing a few smaller issues in mne. :slight_smile:

2 Likes
4

Hi Dan,
Yes I have recently started using mne-python and have gone through the tutorials and examples. I will definitely take a look at the open issues.

Thank you.

1 Like
5

Hi Mikolaj,

Thank for pointing out braindecode. I will look at the ICLabel network and smaller issues as well.

Thanks.

1 Like
6

@mmagnuski A lot of the groundwork has been done by Jacob Feitelberg and Adam Li. See: Python Implementation · Issue #14 · sccn/ICLabel · GitHub and GitHub - mne-tools/mne-icalabel: Automatic labeling of ICA components in Python.
I will also contribute to this, Campus Biotech Geneva has opened and filled an internship position for an ML master student that can work full time on this starting on 1st of March.

1 Like
7

Great, thanks for letting me know @mscheltienne!