Neuroscientific Interpretability
Yeo Seven-Network Functional Brain Mapping
Aggregating Schaefer-100 regional model weights (wkj) into canonical large-scale Yeo 7 functional brain networks to explain task state decoding.
Yeo-7 Network Discriminative Importance (Mk,n)
Mean absolute coefficient weight mapped per functional network
Network Importance Matrix (Mk,n)
Mean absolute coefficient weight per Yeo network
| Yeo Network | Math | Language | Motor | Visual |
|---|---|---|---|---|
| Visual | 0.12 | 0.24 | 0.14 | 0.68 |
| Somatomotor | 0.08 | 0.07 | 0.62 | 0.14 |
| Dorsal Attention | 0.28 | 0.15 | 0.10 | 0.22 |
| Ventral Attn / Salience | 0.22 | 0.20 | 0.21 | 0.09 |
| Limbic | 0.05 | 0.06 | 0.05 | 0.04 |
| Frontoparietal Control | 0.35 | 0.26 | 0.12 | 0.10 |
| Default Mode | 0.18 | 0.34 | 0.08 | 0.06 |
Working Memory / Math
ExecutiveRelied heavily on Frontoparietal Control (0.35), Dorsal Attention (0.28), and Ventral Attention (0.22) networks.
Language / Reading
SemanticBroad involvement across Default Mode (0.34), Control (0.26), and Visual (0.24) networks.
Motor Control
LocalizedOverwhelmingly dominated by the Somatomotor Network (0.62) matching left-vs-right button pressing.
Visual Processing
SensoryDominated by the Visual Network (0.68) with secondary contributions from Dorsal Attention (0.22).