The Geometric Deep Learning core technology cluster consists of PhD students interested in the intersection between geometry, topology and machine learning, and to apply geometrically informed models (e.g. equivariant models) in applications. Subtopics include Group Equivariant Neural Networks, Graph Neural Networks, Topological Data Analysis and more.
For prospective members
To join the cluster, please register via via this link and make sure that you are part of the #ctc_geometric_deep_learning slack channel in the wasp-sweden slack. If you have any problems or questions please send an email to the cluster leader, see contact details below.
Cluster Leader
Cluster Activities
Monthly meetings are held over zoom. Dates, times and links can usually be found in our slack channel about a week before every meeting. If you want to join the meeting, please sign up to the cluster, join the slack channel and/or send an email to the cluster leader.
- January 2025 – Meeting at the WASP Winter Conference, including invited talk by Pavló Melnyk
- December 2024 – Members presentations by Emma Andersdotter (Umeå) and Oscar Carlsson (Chalmers)
- October 2024 – Paper discussion on WL graph networks (paper 1, paper 2)
- September 2024 – Paper discussion Zhdanov et. al. – Implicit Convolutional Kernels for Steerable CNNs
- June 2024 – Presentation discussion at the GeUmetric Deep Learning Workshop at Umeå University
- May 2024 – Paper discussion Brehmer et al. – Geometric Algebra Transformer
- April 2024 – Paper discussion Romero et al. – Attentive Group Equivariant Convolutional Networks
- March 2024 – Paper discussion Cohen et al. – Spherical CNNs
- February 2024 – Discussion meeting, election of new cluster leader
- January 2024 – WASP Winter Conference: Invited speaker Giovanni Marchetti
- December 2023 – Paper discussion Cohen & Welling – Transformation Properties of Learned Visual Representations
- October 2023 – Paper discussion Hutchinson et al. – LieTransformer
- September 2023 – Research presentation Bökman & Kahl – Investigating how ReLU networks encode symmetries
- May 2023 – Study trip to Switzerland.
- March 2023 – Study trip planning
- February 2023 – Minsky & Papert “Perceptrons” book, up until 2.3 “The Group-Invariance Theorem”.
- Spring 2023 – We had weekly meetings going through the UvA Group equivariant DL course: uvagedl.github.io
- January 2023 – WASP Winter Conference
- December 2022 – Paper discussion Batzner et al. – E(3)-equivariant graph neural networks for data-efficient and accurate interatomic potentials
- November 2022 – Paper discussion Jumper et al. – Highly accurate protein structure prediction with AlphaFold
- October 2022 – Paper discussion Gruver et al. – The Lie Derivative for Measuring Learned Equivariance
- September 2022 – Paper discussion Wood and Shawe-Taylor – A unifying framework for invariant pattern recognition
- June 2022 – Study trip to Amsterdam.
- April 2022 – Paper discussion Cesa et al. – A Program to Build E(N)-Equivariant Steerable CNNs
- March 2022 – Paper discussion He et al. – Efficient Equivariant Network
- February 2022 – Paper discussion Jaderberg et al. – Spatial Transformer Networks
- January 2022 – Virtual WASP Winter Conference. Short individual presentations.
- November 2021 – Paper discussion Thomas et al. – Tensor field networks: Rotation- and translation-equivariant neural networks for 3D point clouds
- September 2021 – Meta cluster discussion
- August 2021 – Paper discussion Villar et al. – Scalars are universal: Equivariant machine learning, structured like classical physics
- Fall 2021 – During fall we had weekly meetings to go through the GDL video series
- July 2021 – Paper discussion Dey et al. – Group Equivariant Generative Adversarial Networks, Karras et al. – Alias-Free Generative Adversarial Networks
- June 2021 – Research discussion on equivariance
- May 2021 – Paper discussion Finzi et al. – A Practical Method for Constructing Equivariant Multilayer Perceptrons for Arbitrary Matrix Groups
- April 2021 – Discussing cluster structure
- March 2021 – Intro meeting
Current Members
Below is a list of some of the Geometric Deep Learning cluster members along with short descriptions of their research interests. If you are a cluster member and would like to add your name to the list, please contact the cluster leader. Similarly, if you would like your name to be removed from the list or your info to be edited, also contact the cluster leader.
 |
I am interested in the general study of classes of group- and gauge equivariant neural networks, with emphasis on tools and structures from mathematics and physics. Current projects include the classification of the class of group equivariant transformer models. |
| Affiliation | Mathematical Sciences at Chalmers University of Technology |
| Main Supervisor | Daniel Persson |
| Contact | eliasny@chalmers.se |
 |
Equivariant neural networks and their application to computer vision problems.
|
| Affiliation | Electrical Engineering, Chalmers |
| Main Supervisor | Fredrik Kahl |
| Contact | bokman@chalmers.se, https://www.chalmers.se/en/Staff/Pages/bokman.aspx |
 |
Conformal embedding and neural networks, Models equivariant to geometric transformations
|
| Affiliation | Computer Vision Laboratory, Department of Electrical Engineering, Linköping University |
| Main Supervisor | Michael Felsberg |
| Contact | pavlo.melnyk@liu.se |
 |
My research is focused on investigating and extending the mathematical foundations of equivariant neural networks, with emphasis on Group-equivariant Convolutional Neural Networks (GCNN). Equivariant neural networks are designed to utilize symmetry, when present, to facilitate learning. |
| Affiliation | Mathematical Sciences at Chalmers University of Technology |
| Main Supervisor | Daniel Persson |
| Contact | jimmyar@chalmers.se |
 |
I’m a Ph.D. student in the division of Biomedical Imaging with affiliation to the WASP-AI program. My research’s primary interest is the development and investigation of novel Machine-Learning techniques in biomedical image processing. Specifically, I want to exploit geometrical properties and symmetries of the given image-signals to tackle data-sparsity problems with domain-specific model-design (e.g., Geometrical Deep learning, Graph-CNN, Topology). |
| Affiliation | KTH, Biomedical Imaging |
| Main Supervisor | Rodrigo Moreno |
| Contact | fabiansi@kth.se |