Free Information Center
A steerable convolutional neural network (Steerable CNN) is a type of neural network architecture designed to be equivariant to transformations such as rotations and reflections. By leveraging steerable filters, these networks improve generalization and efficiency in tasks involving geometric variability.
Group-equivariant convolutional neural networks (G-CNNs) are a class of neural networks designed to maintain equivariance to transformations from symmetry groups, enhancing their ability to generalize across transformed inputs. They extend traditional CNNs by incorporating group theory, enabling more efficient learning of invariant features.
Equivariant neural networks are a class of neural network architectures designed to respect symmetry transformations of input data. These networks maintain equivariance under group actions, meaning that transformations applied to inputs correspond predictably to transformations in the output, enabling more efficient learning and better generalization for certain tasks.
SE(3)-equivariant networks are neural network architectures designed to respect the symmetries of the special Euclidean group SE(3), which combines 3D rotations and translations. These networks maintain equivariance under transformations in three-dimensional space, making them particularly useful in fields such as robotics, computer vision, and molecular modeling.
An SO(3)-equivariant neural network is a type of neural network architecture designed to maintain equivariance with respect to the special orthogonal group SO(3), which corresponds to three-dimensional rotations. These networks are used primarily in tasks involving 3D data where rotational symmetry is important, such as molecular modeling, computer vision, and physics simulations.