subspace_overlap¶
- behavenet.fitting.losses.subspace_overlap(A, B, C=None)[source]¶
Compute inner product between subspaces defined by matrices A and B.
- Parameters:
A (
torch.Tensor) – shape (a, d)B (
torch.Tensor) – shape (b, d)C (
torch.Tensor, optional) – shape (c, d)
- Returns:
scalar value; Frobenious norm of UU^T divided by number of entries
- Return type:
torch.Tensor