Category: Activation
Short description: Mish is a Self Regularized Non-Monotonic Neural Activation Function.
Detailed description: Mish is a self regularized non-monotonic neural activation function proposed in the article.
Attributes: operation has no attributes.
Inputs:
Outputs:
Types
Mathematical Formulation
For each element from the input tensor calculates corresponding element in the output tensor with the following formula:
\[ Mish(x) = x*tanh(ln(1.0+e^{x})) \]
Examples