Manifolds Package

The module implements a variety of Riemannian manifolds as Python classes. Each manifold class encapsulates core geometric operations, such as distance functions, geodesics, exponential/logarithm maps, projections, and retractions, enabling integration of non-Euclidean geometry into neural network models.

Certain manifolds (e.g., SPD and Stiefel) can be endowed with different Riemannian metrics, resulting in distinct operations implemented in their corresponding classes.