Clustering
Description:
Clustering, i.e., grouping elements based on similar features, is a fundamental task in machine learning. We develop fast algorithms for the multicut formulation of the clustering problem. A special focus is placed on scalable, theoretically grounded algorithms that operate on datasets that previously required too long runtimes. In this context, we are interested in GPU solvers. Another focus is on dense multicut, i.e., clustering problems where the similarity features are given for all pairs of points.
Literatur:
- "ClusterFuG: Clustering Fully connected Graphs by Multicut", Abbas, Ahmed and Swoboda, Paul, IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 2023
- "RAMA: A Rapid Multicut Algorithm on GPU", Abbas, Ahmed and Swoboda, Paul, Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 2022
- "Combinatorial persistency criteria for multicut and max-cut", Lange, Jan-Hendrik and Andres, Bjoern and Swoboda, Paul, Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR)2019
- "A message passing algorithm for the minimum cost multicut problem", Swoboda, Paul and Andres, Bjoern, Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 2017