BEGIN:VCALENDAR VERSION:2.0 PRODID:researchseminars.org CALSCALE:GREGORIAN X-WR-CALNAME:researchseminars.org BEGIN:VEVENT SUMMARY:Mathilde Papillon (UCSB - USA) DTSTART:20250131T160000Z DTEND:20250131T170000Z DTSTAMP:20260423T041527Z UID:GEOTOP-A/94 DESCRIPTION:Title: Make any GNN Go Topological with TopoTune\nby Mathilde Papillon (UCS B - USA) as part of GEOTOP-A seminar\n\n\nAbstract\nGraph Neural Networks (GNNs) excel in learning from relational datasets\, processing node and ed ge features in a way that preserves the symmetries of the graph domain. Ho wever\, many complex systems--such as biological or social networks--invol ve multiway complex interactions that are more naturally represented by hi gher-order topological spaces. The emerging field of Topological Deep Lear ning (TDL) aims to accommodate and leverage these higher-order structures. Combinatorial Complex Neural Networks (CCNNs)\, fairly general TDL models \, have been shown to be more expressive and better performing than GNNs. However\, differently from the graph deep learning ecosystem\, TDL lacks a principled and standardized framework for easily defining new architectur es\, restricting its accessibility and applicability. To address this issu e\, we introduce in this talk Generalized CCNNs (GCCNs)\, a novel simple y et powerful family of TDL models that can be used to systematically transf orm any (graph) neural network into its TDL counterpart. We show how GCCNs generalize and subsume CCNNs\, and briefly describe extensive experiments on a diverse class of GCCNs. We show that these architectures consistentl y match or outperform CCNNs\, often with less model complexity. In an effo rt to accelerate and democratize TDL\, we also introduce TopoTune\, a ligh tweight software that allows practitioners to define\, build\, and train G CCNs with unprecedented flexibility and ease.\n LOCATION:https://researchseminars.org/talk/GEOTOP-A/94/ END:VEVENT END:VCALENDAR