The article presents Clifford-Steerable Convolutional Neural Networks (CS-CNNs), a new class of E(p, q)-equivariant CNNs. These networks process multivector fields on pseudo-Euclidean spaces Rp,q, offering significant improvements over baseline methods in fluid dynamics and electrodynamics forecasting tasks. The CS-CNNs are the first models to respect the full spacetime symmetries of these problems, providing consistently better performance across different dataset sizes. The work is a significant contribution to the field of deep learning and neural networks.


Publication date: 23 Feb 2024
Project Page: Not Provided