Missed TensorFlow Dev Summit? Check out the video playlist. Watch recordings


TensorFlow 1 version View source on GitHub

Computes log of the determinant of a hermitian positive definite matrix.

    matrix, name=None
# Compute the determinant of a matrix while reducing the chance of over- or
A = ... # shape 10 x 10
det = tf.exp(tf.linalg.logdet(A))  # scalar


  • 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.


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.