cholesky_invert#
- hypercoil.functional.matrix.cholesky_invert(X: Tensor) Tensor[source]#
Invert a symmetric positive definite matrix using Cholesky decomposition.
Warning
The input matrix must be symmetric and positive definite. If this is not the case, the function will either raise a LinAlgError or produce undefined results. For positive semidefinite matrices, the Moore-Penrose pseudoinverse can be used instead.
Performance
This does not appear to be any faster than using the inverse directly. In fact, it is almost always slower than
jnp.linalg.inv. It’s retained for historical reasons.- Dimension:
- Input : \((*, D, D)\)
D denotes the row or column dimension of the matrices to be inverted.
*denotes any number of preceding dimensions.- Output : \((*, D, D)\)
As above.
- Parameters:
- ATensor
Symmetric positive definite matrix.
- Returns:
- AinvTensor
Inverse of the input matrix.