Figure 3.

Visualization of LDA-driven embeddings for simulated data where the number of nodes varies p ∈ {20, 50, 100, 150}. Each panel shows the mean LDA-driven embedding over 10 unseen, simulated datasets (i.e., 10 unseen subjects). The thick blue line corresponds to the average of the linear discriminant scores across subjects. Standard deviations are indicated by the shaded regions. The dashed red lines show the linear discriminant scores for three randomly selected subjects, thereby providing an indicating of the variability across subjects. Results are shown when the underlying connectivity structure was simulated using three distinct graph algorithms: Erdős-Rényi, scale-free and small-world random graphs. Vertical dashed lines indicate a change in covariance structure.