<ul><li>Semi-Simplicial Neural Networks (SSNs) are introduced to capture directed higher-order motifs and their directional relationships in complex systems like brain networks.</li><li>Routing-SSNs dynamically select informative relations to enhance scalability, outperforming existing models in expressiveness and performance.</li><li>SSNs are proven to be more expressive than standard graph and Topological Deep Learning (TDL) models, leading to state-of-the-art performance in brain dynamics classification tasks.</li><li>The study establishes the potential of topological models like SSNs for learning from structured brain data, with the results demonstrating competitive performance on node classification and edge regression tasks as well.</li></ul>

Directed Semi-Simplicial Learning with Applications to Brain Activity Decoding

Discover more