Type Softsign
Namespace tensorflow.contrib.distributions.bijectors
Parent Bijector
Interfaces ISoftsign
Bijector which computes `Y = g(X) = X / (1 + |X|)`. The softsign `Bijector` has the following two useful properties: * The domain is all real numbers
* `softsign(x) approx sgn(x)`, for large `|x|`. #### Examples
Show Example
# Create the Y = softsign(X) transform.
softsign = Softsign()
x = [[[1., 2],
[3, 4]],
[[5, 6],
[7, 8]]]
x / (1 + abs(x)) == softsign.forward(x)
x / (1 - abs(x)) == softsign.inverse(x)