AMLab | Amsterdam Machine Learning Lab

Georg Bökman

Postdoc
AMLab
Institute of Informatics
University of Amsterdam
Science Park, Lab 42, L4.7

Personal page Google scholar Github

I am a postdoc working on geometric deep learning, hosted by Erik Bekkers on a WASP postdoc scholarship. I did my PhD at Chalmers under the supervision of Fredrik Kahl.

Two things I’m interested in are 1) making geometric deep learning methods efficient and 2) the practical and theoretical tradeoffs between learning a priori known problem symmetries from data versus constraining the learning model to respect the symmetries.

Selected Publications

ICML

Flopping for FLOPs: Leveraging Equivariance for Computational Efficiency

Bökman, Georg, Nordström, David, and Kahl, Fredrik

In Forty-second International Conference on Machine Learning Dec 2025

Abs Code

Incorporating geometric invariance into neural networks enhances parameter efficiency but typically increases computational costs. This paper introduces new equivariant neural networks that preserve symmetry while maintaining a comparable number of floating-point operations (FLOPs) per parameter to standard non-equivariant networks. We focus on horizontal mirroring (flopping) invariance, common in many computer vision tasks. The main idea is to parametrize the feature spaces in terms of mirror-symmetric and mirror-antisymmetric features, i.e., irreps of the flopping group. This decomposes the linear layers to be block-diagonal, requiring half the number of FLOPs. Our approach reduces both FLOPs and wall-clock time, providing a practical solution for efficient, scalable symmetry-aware architectures.
CVPR

Steerers: A framework for rotation equivariant keypoint descriptors

Bökman, Georg, Edstedt, Johan, Felsberg, Michael, and Kahl, Fredrik

In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Dec 2024

Abs Code

Image keypoint descriptions that are discriminative and matchable over large changes in viewpoint are vital for 3D reconstruction. However, descriptions output by learned descriptors are typically not robust to camera rotation. While they can be made more robust by, e.g., data augmentation, this degrades performance on upright images. Another approach is test-time augmentation, which incurs a significant increase in runtime. Instead, we learn a linear transform in description space that encodes rotations of the input image. We call this linear transform a steerer since it allows us to transform the descriptions as if the image was rotated. From representation theory, we know all possible steerers for the rotation group. Steerers can be optimized (A) given a fixed descriptor, (B) jointly with a descriptor or (C) we can optimize a descriptor given a fixed steerer. We perform experiments in these three settings and obtain state-of-the-art results on the rotation invariant image matching benchmarks AIMS and Roto-360.