 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.