tf.linalg.logdet
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.
© 2020 The TensorFlow Authors. All rights reserved.
Licensed under the Creative Commons Attribution License 3.0.
Code samples licensed under the Apache 2.0 License.
https://www.tensorflow.org/versions/r1.15/api_docs/python/tf/linalg/logdet