Joseph Paillard

I could reproduce the issue: the model doesn't learn anything, the accuracy is constant and equal to 0.2. I solved it by parsing the argument `picks=['Fpz-Cz', 'Pz-Oz']` to `create_windows_from_events()`. Then...

Hi ! Interesting implementation, thx for sharing. I was wondering if there is a reason why you adopted Riemannian gradients with geoopt instead of a parameterized optimization approach such as...