det_op

→TensorImplPtr

int det_op(const int & a)

source edit

Python wrappers

pylucid.linalg.detfunction

Compute $\det(A)$ for a square float tensor.

Validates that a is at least 2-D, square in the trailing two dimensions, and float-typed; dispatches to the backend; and wires DetBackward into the autograd graph (saving both $A$ and the scalar output) when grad mode is active.

Shape

a: (..., N, N).
Output: (...,) (rank decreases by 2; a single 2-D input yields a 0-D scalar).

See Also

inv_op : Matrix inverse, used in the backward.

Parameters

aTensorImplPtr

Input matrix of shape (..., N, N).

Returns

TensorImplPtr

Scalar-per-batch determinant of shape (...,) (the trailing two matrix dimensions are removed).

Raises

ValueError

If a is fewer than 2 dimensions, non-square, or non-float.

LinAlgError

(At backend level.) If

A

is singular — LU returns a zero pivot.

Notes

For ill-conditioned $A$ prefer slogdet (sign + log) when it becomes available; the magnitude of det can over/underflow long before $A$ is numerically singular.