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.
    View source on GitHub