e3x.nn.activations.selu

e3x.nn.activations.selu(x)[source]

Scaled exponential linear unit activation.

Computes the gated linear activation with:

\[\begin{split}\mathrm{gate}(x) = \lambda \begin{cases} 1, & x > 0\\ \frac{\alpha}{x} \left(\exp(x) - 1\right), & x \le 0 \end{cases}\end{split}\]

where \(\lambda = 1.0507009873554804934193349852946\) and \(\alpha = 1.6732632423543772848170429916717\).

For scalar inputs, this is equivalent to:

\[\begin{split}\mathrm{selu}(x) = \lambda \begin{cases} x, & x > 0\\ \alpha \left(\exp(x) - 1\right), & x \le 0 \end{cases}\end{split}\]

For more information, see Self-Normalizing Neural Networks.

../_images/e3x.nn.activations.selu_0_0.svg
Parameters:

x (Union[Float[Array, '... 1 (max_degree+1)**2 num_features'], Float[Array, '... 2 (max_degree+1)**2 num_features']]) – Input features to which the nonlinearity is applied.

Return type:

Union[Float[Array, '... 1 (max_degree+1)**2 num_features'], Float[Array, '... 2 (max_degree+1)**2 num_features']]

Returns:

The result of applying the nonlinearity to the input features.