[EHN] EEGMiner

[bruAristimunha]

Small benchmark

Cross-Session

Cross-Subject:

[bruAristimunha]

I am planning only to include tests

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 86.90%. Comparing base (c316bcc) to head (e9cd97b).
Report is 2 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #667      +/-   ##
==========================================
+ Coverage   86.58%   86.90%   +0.31%     
==========================================
  Files          72       73       +1     
  Lines        6450     6604     +154     
==========================================
+ Hits         5585     5739     +154     
  Misses        865      865              

[sylvchev]

@mccorsi it could be interesting to see if others functional connectivity could be used with this type of neural architecture

[sylvchev]

Regarding the licence, you could add a NOTICE file to indicate that EEGMiner is CC by-nc

[mccorsi]

Hi @sylvchev, do have constraints on the FC matrices you want to use (ie should they have to be SPD...)? I might have some ideas but I'd like to be sure that it makes sense for the project! :)

[bruAristimunha]

No constraints on the FC @mccorsi and @sylvchev

[sylvchev]

Nice work. Thanks @bruAristimunha !

[mccorsi]

Here is a non-exhaustive list of FC metrics :

• Coherence-based metrics: cf the one already implemented in pyriemann & used in FUCONE
• more generally, mne-python increased the set of spectral FC metrics: Coherence, Coherency, Imaginary part of Coherency, Canonical Coherency (CaCoh), Maximised Imaginary part of Coherency (MIC), Multivariate Interaction Measure (MIM), Phase-Locking Value (PLV), Corrected Imaginary PLV (ciPLV), Pairwise Phase Consistency (PPC), Phase Lag Index (PLI), Unbiased estimator of squared PLI, Directed PLI (DPLI), Weighted PLI (WPLI), Debiased estimator of squared WPLI, State-space Granger Causality (GC), State-space GC on time-reversed signals
• bicoherence: to measure the coherence from two different bins of frequency

I am currently working on metrics that would capture non-linearities but it is still under development. I will let you know when it is ready (and implemented in python)!