Code & Resources: https://github.com/lkorczowski/BCI-2021-Riemannian-Geometry-workshop/

Please give us feedback on your satisfaction of the workshop: https://iyrp0g0k85t.typeform.com/to/g6Mcub4o

If you have more questions, you can add them below.

(talk) How Riemannian Geometry transformed BCI? - Marco Congedo

slides

• Do you know if this method can applied to EMG data?
• Could RPA work in an online setting? A trivial approach would be, after each new data point, recompute entirely from scratch, but this will be costly. Is there a better approach?
• Can Riemannian Geometry be used for P300 detection?
• Could you provide more insights for constructing covariance matrix for ERP instead of the super trial method
• Are there downsides of translating back to the tangent space vs. working in he RG space directly?

(talk) Why Riemannian Geometry works so well? - Florian Yger

slides

• Is there any way to compute the mass of class without using gradient descent?
• How do you define an equivalent cost function on tangent space to a cost function on the manifold?
• Do other non-Euclidean geometries also perform well on these problems or are the benefits exclusive to Riemannian geometry?
• Are you aware of an example where the swelling effect negatively affects the performance of an ML method in practice?