---
title: Clifford-Steerable CNNs
url: https://www.emergentmind.com/topics/clifford-steerable-cnns
type: topic
---

# Clifford-Steerable CNNs

Clifford-Steerable Convolutional Neural Networks (CS-CNNs) are a class of deep learning architectures designed to achieve exact equivariance under pseudo-Euclidean isometry groups, unifying the treatment of translational, orthogonal, and Lorentzian symmetries within a single convolutional framework. These networks process feature fields whose values are multivectors in Clifford algebras, allowing natural and explicit representations for a wide range of physical quantities. CS-CNNs implement implicit, group-steerable convolutional kernels via Clifford group-equivariant neural networks, and have demonstrated state-of-the-art performance on partial differential equation (PDE) forecasting tasks including fluid dynamics and relativistic electrodynamics [2402.14730][2510.14007].

## 1. Pseudo-Euclidean Symmetry and Clifford Algebra

A central aspect of CS-CNNs is their principled equivariance to the isometry group $E(p,q)=\mathbb{R}^{p+q}\ltimes O(p,q)$, where $O(p,q)$ acts as the pseudo-orthogonal (Lorentz) group on $\mathbb{R}^{p,q}$, a vector space with metric signature $(p,q)$; for example, $(3,0)$ for classical Euclidean space and $(1,3)$ for Minkowski spacetime. 

The Clifford algebra $\mathrm{Cl}(p,q)$, generated by orthogonalizing basis vectors $e_i$ under the bilinear form $\eta^{p,q}$, provides the representation space for the multivector-valued feature fields directly encoding scalar, vector, and higher-grade quantities. The $O(p,q)$ action is lifted to the entire algebra via the natural grading-preserving algebra automorphism $\rho_{\text{cliff}}$, enabling precise equivariant transformations for all geometric components present in the data [2402.14730][2510.14007].

## 2. Equivariant Convolutions and Steerable Kernel Parameterization

A convolutional layer $L$ is $E(p,q)$-equivariant if and only if its kernel $K:\mathbb{R}^{p,q}\to\text{End}(\mathrm{Cl}(p,q)^c)$ satisfies the steerability relation:
$$
K(gx) = \rho_{\text{out}}(g)\, K(x)\, \rho_{\text{in}}(g)^{-1} \quad \forall g\in O(p,q),\, x\in\mathbb{R}^{p,q}.
$$

CS-CNNs realize such steerable kernels via an implicit parameterization. Specifically, a Clifford group-equivariant neural network (CGENN) $K_0:\mathbb{R}^{p,q}\to\mathrm{Cl}(p,q)^{c\times c}$ produces c-channel multivector outputs. These are passed through a fixed, linear "kernel head" $H$ that expands multivector entries into $\mathbb{R}$-linear endomorphisms of $\mathrm{Cl}(p,q)^c$ by weighted, partial geometric product expansions, with learnable mixing weights $\lambda_{mn}^{k,ij}$:
$$
H(K_0)[f]_i^{(k)} = \sum_{j=1}^c \sum_{m,n=0}^d \lambda_{mn}^{k,ij} \left( K_{0,ij}^{(m)} \cdot f_j^{(n)} \right)^{(k)}.
$$

Together, equivariance of $K_0$ and $H$ ensures that their composition $K=H\circ K_0$ satisfies the group steerability constraint [2402.14730].

## 3. Network Architecture, Implementation, and Optimization

The principal components of CS-CNNs include:

- **Input/Output:** Channel-wise multivector fields $f(x)\in\mathrm{Cl}(p,q)^c$ decomposed into grades.
- **Steerable convolution:** $(f\ast K)(u) = \int f(y)\cdot K(y^{-1}u) \, dy$.
- **Nonlinearity:** Grade-wise geometric-product gating, e.g., $A(x) = x \cdot \sigma(x^{(0)})$ with a scalar nonlinearity $\sigma$ on the grade-0 part.
- **Residual blocks:** Combination of steerable convolution, group normalization, activation, and skip connections.

Optimization techniques include pre-computation of geometric-product Cayley tables for efficient kernel expansion, the use of FFT-based (Fast Fourier Transform) convolution and backpropagation, and the implementation of CGENNs with grade-structured linear and geometric-product layers [2402.14730].

## 4. Kernel Basis Completeness and Conditional Kernels

The standard CS-CNN parameterization does not realize a complete basis of equivariant steerable kernels. For example, in the O(2)-equivariant vector-to-vector setting, harmonic components at frequency two (cos$2\varphi$, sin$2\varphi$) are missing from the representable space when kernels depend solely on the spatial offset $z$, limiting expressivity.

Conditional Clifford-Steerable CNNs (C-CSCNNs) solve this by incorporating global, translation-invariant summaries $T[f]$ of the input field as additional kernel arguments:
$$
f_\text{out}(x) = \int \hat{K}(x-y,\,T[f])\, f(y)\, dy.
$$
Here, $T[f]$ is typically the mean pooled multivector over spatial regions, ensuring both translation and $O(p,q)$-equivariance. The conditional kernel $\hat{K}$ is parameterized via an O($p,q$)-equivariant network $\hat{\mathcal{K}}(z, T[f])$ and the kernel head $H$ as before, restoring the missing harmonics and the full kernel basis in a single layer [2510.14007].

## 5. Empirical Performance in PDE Modeling

Empirical evaluations demonstrate that CS-CNNs and their conditional variants achieve robust, data-efficient results on diverse PDE forecasting tasks including Navier–Stokes (R$^2$), shallow-water equations, non-relativistic and relativistic Maxwell equations in both Euclidean and Minkowski space:

| Task              | Metric                  | Best C-CSCNN Result                                     | Comparison    |
|-------------------|------------------------|---------------------------------------------------------|--------------|
| Navier–Stokes     | 1-step MSE             | C-CSCNN halves MSE of plain CS-CNN with 512 trajectories | Outperforms ResNet, Transolver, Swin-Transformer, FNO |
| Shallow Water SWE | 5-step relative L$^2$  | C-CSCNN-L: 2.94% (55M params); C-CSCNN-S: 3.51% (10M)   | Better or competitive with much larger baselines       |
| Maxwell (3D)      | 1-step field MSE       | C-CSCNN halves error vs. CS-CNN and outperforms FNO/G-FNO| -            |
| Maxwell (R$^{1,2}$)| 1-stepfield MSE       | Similar data-efficiency gains across data regimes        | -            |

In all tasks, both CSCNNs and C-CSCNNs maintain relative equivariance error $\mathcal{O}(10^{-7})$ and error maps demonstrate lower transient and stable long-horizon errors compared to baselines. Notably, C-CSCNNs achieve improved data efficiency, enabling smaller models to match or surpass much larger transformer- and FNO-based architectures [2510.14007][2402.14730].

## 6. Limitations, Computational Cost, and Extensions

CS-CNNs currently support only full multivector (Cl($p,q$)) representations and do not directly realize all $O(p,q)$ irreducible representations (e.g., spinors). Some kernel degrees of freedom require stacking multiple layers due to the implicit parameterization, and practical implementation on noncompact $O(p,q)$ groups necessitates grid truncation, leading to approximate equivariance under large boosts.

Feature dimensionality scales as $2^d$ per channel (with $d=p+q$), and geometric-product layers are computationally more expensive than conventional ReLU nonlinearities, though tractable up to $d=4$. Prospective extensions include applying alternative Clifford-based nonlinearities, conditioning at local or hierarchical levels, adaptations to pseudo-Riemannian manifolds, and the construction of physics-informed architectures, e.g., using steerable PDE operators. Conditioning on global or learned multivector summaries further improves the kernel basis, enabling stable and expressive modeling of a broader class of physical phenomena [2510.14007][2402.14730].

## 7. Impact and Future Research Directions

Clifford-Steerable CNNs provide a unified, E($p,q$)-equivariant deep learning framework grounded in Clifford algebra and group theory, enabling exact symmetry-preserving modeling of fields with diverse geometric content across varied pseudo-Euclidean spaces. Conditional kernel mechanisms resolve basis incompleteness while maintaining computational tractability, expanding applicability to challenging PDEs and physical forecasting problems with notable gains in accuracy, efficiency, and long-term stability.

Possible future research includes: local/hierarchical or adaptive conditioning, expansion to curved manifold scenarios via parallel-transport convolution, and application domains such as molecular dynamics, geophysical flows, and high-energy physics on space-times of different signatures. This suggests continued investigation of Clifford-algebraic neural architectures and generalized symmetry-informed networks for scientific machine learning [2510.14007][2402.14730].

Source: https://www.emergentmind.com/topics/clifford-steerable-cnns