Figure 6
From: Geometric learning of functional brain network on the correlation manifold
Embeddings onto 2-dimensional Euclidean space for a collection of motor imagery correlation matrices. Rows correspond to a choice of geometric structure by viewing the matrices as elements of the correlation manifold (top), the SPD manifold (middle), and the Euclidean space (bottom). Columns represent algorithms used under each geometry, including classical multidimensional scaling (left), metric multidimensional scaling (center), and principal geodesic analysis (right). Colors represent two classes of fists (in red) and feet (in blue).
