Topological Blind Spots: Understanding and Extending Topological Deep Learning Through the Lens of Expressivity

• Published in: arXiv.org
• Main Authors: Yam Eitan, Gelberg, Yoav, Bar-Shalom, Guy, Frasca, Fabrizio, Bronstein, Michael, Maron, Haggai
• Format: Paper
• Language: English
• Published: Ithaca Cornell University Library, arXiv.org 10.08.2024
• Subjects: Cellular structure; Datasets; Deep learning; Graphical representations; Homology; Message passing; Neural networks; Synthetic data; Topology

Topological deep learning (TDL) facilitates learning from data represented by topological structures. The primary model utilized in this setting is higher-order message-passing (HOMP), which extends traditional graph message-passing neural networks (MPNN) to diverse topological domains. Given the si...