Table 1 Average computational time (in s) to compute Fréchet mean of N correlation matrices of size \((n\times n)\) under correlation and SPD geometry.
From: Geometric learning of functional brain network on the correlation manifold
\(n=5\) | \(n=10\) | \(n=15\) | \(n=30\) | ||
|---|---|---|---|---|---|
\(N=20\) | CORR | 0.183 | 0.487 | 0.839 | 2.249 |
SPD | 0.008 | 0.016 | 0.035 | 0.073 | |
\(N=50\) | CORR | 0.383 | 0.988 | 1.690 | 5.365 |
SPD | 0.017 | 0.032 | 0.069 | 0.171 | |
\(N=100\) | CORR | 0.813 | 2.132 | 3.387 | 10.743 |
SPD | 0.035 | 0.066 | 0.145 | 0.341 | |
