Group-Equivariant Poincaré Convolutional Networks
For researchers in geometric deep learning and computer vision, this work addresses the challenge of incorporating equivariance into hyperbolic networks, which previously treated spatial transformations as distinct concepts.
The paper tackles the problem of learning visual representations in hyperbolic space with group equivariance, proposing Equivariant Poincaré ResNets that combine hyperbolic geometry with discrete symmetry groups. The method achieves faster convergence and reduced parameter usage while preserving spatial-group equivariance.
While recent advancements like the Poincaré ResNet have demonstrated the potential of learning visual representations directly in hyperbolic space, their optimisation remains hampered by the computationally intensive nature of Riemannian gradients and the strict boundaries of the manifold. Furthermore, standard hyperbolic networks treat spatial transformations of the same object as distinct hierarchical concepts, leading to redundant parameter usage and vanishing signals. We propose Equivariant Poincaré ResNets, combining hyperbolic geometry with discrete symmetry groups ($C_4$ and $D_4$). We identify critical roadblocks in applying Euclidean equivariance to hyperbolic space and propose geometrically safe tensor reshaping, left-regular permutations for hyperbolic group convolutions, and joint-orientation Poincaré Midpoint Batch normalisation. Empirically, embedding equivariance drastically reduces the optimisation space, accelerating convergence while accelerating convergence while respecting the boundary constraints of the Poincaré ball and preserving spatial-group equivariance.