spd_learn.functional.pcm_geodesic#

spd_learn.functional.pcm_geodesic(A: Tensor, B: Tensor, t: float | Tensor, theta: float = 0.5) → Tensor[source]#

Compute geodesic interpolation under the Power-Cholesky Metric (PCM).

Parameters:

  • A (torch.Tensor) – Starting and ending SPD matrices of shape (…, n, n).

  • B (torch.Tensor) – Starting and ending SPD matrices of shape (…, n, n).

  • t (float or torch.Tensor) – Interpolation parameter \(t \in [0, 1]\).

  • theta (float, default=0.5) – Power parameter (\(\theta \neq 0\)).

Returns:

Interpolated SPD matrix of shape (…, n, n).

Return type:

torch.Tensor