Authors: Maksim Zhdanov, David Ruhe, Maurice Weiler, and Ana Lucic, Johannes Brandstetter, Patrick Forré
We present Clifford-Steerable Convolutional Neural Networks (CS-CNNs), a novel class of
To install all the necessary requirements, including JAX and PyTorch (CPU), run:
chmod +x setup.sh
./setup.sh
The repository is incomplete at the moment, below is the roadmap:
- implementation of Clifford-steerable kernels/convolutions (in JAX)
- implementation of Clifford-steerable ResNet and basic ResNet (in JAX)
- demonstrating example + test equivariance (escnn + PyTorch required)
- implementation of Clifford ResNet and Steerable ResNet (in PyTorch)
- code for the data generation (Maxwell on spacetime)
- replicating experimental results
If you find this repository useful in your research, please consider citing us:
@misc{zhdanov2024cliffordsteerable,
title={Clifford-Steerable Convolutional Neural Networks},
author={Maksim Zhdanov and David Ruhe and Maurice Weiler and Ana Lucic and Johannes Brandstetter and Patrick Forré},
year={2024},
eprint={2402.14730},
archivePrefix={arXiv},
primaryClass={cs.LG}
}