At the end of July, Sebastian Mežnar, Boshko Koloski, and Blaž Škrlj participated in the “ICML Topological Deep Learning Challenge 2024: Beyond the Graph Domain” at the ICML 2024 conference, where they secured third place in the Best Feature-Based Lifting category, with the paper titled: “Mixture of Gaussians + MST Lifting (Pointcloud to Hypergraph)”. Their innovative approach involved transforming a point cloud into a hypergraph by employing a Gaussian mixture model and constructing a minimal spanning tree between the generated Gaussians. This challenge was part of the broader “Geometry-grounded Representation Learning and Generative Modeling” workshop.
Congratulations!