View source on GitHub
|
Computes log of the determinant of a hermitian positive definite matrix.
tf.linalg.logdet(
matrix, name=None
)
# Compute the determinant of a matrix while reducing the chance of over- or
underflow:
A = ... # shape 10 x 10
det = tf.exp(tf.linalg.logdet(A)) # scalar
Args | |
|---|---|
matrix
|
A Tensor. Must be float16, float32, float64, complex64,
or complex128 with shape [..., M, M].
|
name
|
A name to give this Op. Defaults to logdet.
|
Returns | |
|---|---|
The natural log of the determinant of matrix.
|
numpy compatibility
Equivalent to numpy.linalg.slogdet, although no sign is returned since only hermitian positive definite matrices are supported.