Figure 3
From: Jumping over baselines with new methods to predict activation maps from resting-state fMRI
Predictive \(R^2\) score benchmark results: scores indicate the mean of cortical surface \(R^2\), see Eq. (1), weighted by the variance of each surface vertex. The colorbar indicates this measure. Roughly half of contrast targets have mean cortical \(R^2\) below 0 since predictive \(R^2\) can be arbitrarily negative. Math-Story stands out as the easiest contrast to predict. A discussion providing a reason why is provided in “Spatially resolved predictability” section. Column and row ordering are not sorted by performance and remains identical to Fig. 2. The left most column is the mean score across all contrasts.
