spd_learn.functional.exp_map_pcm#
- spd_learn.functional.exp_map_pcm(V: Tensor, P: Tensor | None = None, theta: float = 0.5, eps: float = 1e-06) Tensor[source]#
Compute Riemannian exponential map under the Power-Cholesky Metric.
- Parameters:
V (torch.Tensor) – Tangent matrix / deformed coordinates of shape (…, n, n).
P (torch.Tensor, optional) – Base SPD matrix of shape (…, n, n). If None, identity \(I\) is used.
theta (float, default=0.5) – Power parameter (\(\theta \neq 0\)).
eps (float, default=1e-6) – Numerical threshold to clamp diagonal.
- Returns:
Reconstructed SPD matrix of shape (…, n, n).
- Return type: