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.