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


View source on GitHub

Computes Winitzki approximation to Lambert W function at z >= -1/exp(1).

    z, name=None

The approximation for z >= -1/exp(1) will be used as a starting point in the iterative algorithm to compute W(z). See _lambertw_principal_branch() below. See www.researchgate.net/post/Is_there_approximation_to_the_LambertWx_function and in particular (38) in pdfs.semanticscholar.org/e934/24f33e2742016ef18c36a80788400d2f17b4.pdf


  • z: value for which W(z) should be computed. Expected z >= -1/exp(1). If not then function will fail due to log(<0).
  • name: optionally pass name for output.


  • lambertw_winitzki_approx: Approximation for W(z) for z >= -1/exp(1).